I'm try to bond a few interfaces together with the hopes of getting increased throughput, and I'm using a cisco Catalyst 2900 as the switch. I've tried using mode 0, 5, and 6 with nothing special on the switch, and mode 4 with some ports "trunked" together (I have a feeling that the "trunking" that the 2900 does is not 802.3ad, as it disabled the ports it saw as redundant), yet xfer speeds always cap out at about 10MB/s.
Has any body accomplished bonding with increased throughput as the goal, with or without (without might be preferable) doing something special on the switch (preferably the afore-mentioned Catalyst 2900, as that is what I have to work with as a non-sactioned side-project ;)?
--Tim ___________________________________________________ < When pleasure remains, does it remain a pleasure? > --------------------------------------------------- \ \ \ \ /\ ( ) .( o ).
I'm try to bond a few interfaces together with the hopes of getting increased throughput, and I'm using a cisco Catalyst 2900 as the switch. I've tried using mode 0, 5, and 6 with nothing special on the switch, and mode 4 with some ports "trunked" together (I have a feeling that the "trunking" that the 2900 does is not 802.3ad, as it disabled the ports it saw as redundant), yet xfer speeds always cap out at about 10MB/s.
Has any body accomplished bonding with increased throughput as the goal, with or without (without might be preferable) doing something special on the switch (preferably the afore-mentioned Catalyst 2900, as that is what I have to work with as a non-sactioned side-project ;)?
--Tim
I just did this with an HP switch, it was obviously easier:) Have a look here: http://www.mjmwired.net/kernel/Documentation/networking/bonding.txt
Some good info.
HTH, jlc
On Wed, 2008-04-09 at 15:47 -0700, Timothy Selivanow wrote:
I'm try to bond a few interfaces together with the hopes of getting increased throughput, and I'm using a cisco Catalyst 2900 as the switch. I've tried using mode 0, 5, and 6 with nothing special on the switch, and mode 4 with some ports "trunked" together (I have a feeling that the "trunking" that the 2900 does is not 802.3ad, as it disabled the ports it saw as redundant), yet xfer speeds always cap out at about 10MB/s.
Has any body accomplished bonding with increased throughput as the goal, with or without (without might be preferable) doing something special on the switch (preferably the afore-mentioned Catalyst 2900, as that is what I have to work with as a non-sactioned side-project ;)?
IEEE 802.1Q trunking Supported. Cisco IOS Release 11.2(8)SA5 (Enterprise Edition Software)
Inter-Switch Link (ISL) trunking Cisco IOS Release 11.2(8)SA4 (Enterprise Edition Software)
set fastether-options 802.3ad (((try that)))?
I have a Cisco Cert but that does not mean anything. Have not worked on ciso equipment in over 4 years.
--Tim ___________________________________________________
< When pleasure remains, does it remain a pleasure? >
\ \ \ \ /\ ( ) .( o ).
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 4/9/08, Timothy Selivanow timothy.selivanow@virtualxistenz.com wrote:
I'm try to bond a few interfaces together with the hopes of getting increased throughput, and I'm using a cisco Catalyst 2900 as the switch. I've tried using mode 0, 5, and 6 with nothing special on the switch, and mode 4 with some ports "trunked" together (I have a feeling that the "trunking" that the 2900 does is not 802.3ad, as it disabled the ports it saw as redundant), yet xfer speeds always cap out at about 10MB/s.
Has any body accomplished bonding with increased throughput as the goal, with or without (without might be preferable) doing something special on the switch (preferably the afore-mentioned Catalyst 2900, as that is what I have to work with as a non-sactioned side-project ;)?
--Tim ___________________________________________________
< When pleasure remains, does it remain a pleasure? >
\ \ \ \ /\ ( ) .( o ).
Hi there
As another person with cisco certs that aren't being used i wondered about the port being switched off which sounds like a spanning-tree issue.
/me dredging up heavily repressed stuff from the BCMSN
Certainly the 2900 will support 802.3ad or LACP natively.
i found this which may be of use
http://wiki.oracle.com/page/Cisco+Systems+IOS-based+switches-+interface+bonding+and+trunking?t=anon
mike
On Thu, 2008-04-10 at 11:24 +0100, Michael Simpson wrote:
Certainly the 2900 will support 802.3ad or LACP natively.
i found this which may be of use
http://wiki.oracle.com/page/Cisco+Systems+IOS-based+switches-+interface+bonding+and+trunking?t=anon
So, as it turns out, it's a 2900XL, which does not support 802.3ad or LACP at all, just a proprietary port channeling for switch interconnect as far as I can tell.
I suppose that leaves me with just using pure software (for now, I have an 8 port Intel Pro 100 at home that I'll look at...), but I'm unable to get the increased throughput using mode=0. I did notice, however, while I was on the switch console it kept complaining about the interfaces flapping and re-learning addresses. My network guy here at work said that it was bad and either the switch or the bond(s) is misconfigured somewhere. Any hints as to where and or what kinds of things I should be looking at?
--Tim ______________________________________________________ / The truth is what is; what should be is a dirty lie. \ \ -- Lenny Bruce / ------------------------------------------------------ \ \ \ \ /\ ( ) .( o ).
Timothy Selivanow wrote:
On Thu, 2008-04-10 at 11:24 +0100, Michael Simpson wrote:
Certainly the 2900 will support 802.3ad or LACP natively.
i found this which may be of use
http://wiki.oracle.com/page/Cisco+Systems+IOS-based+switches-+interface+bonding+and+trunking?t=anon
So, as it turns out, it's a 2900XL, which does not support 802.3ad or LACP at all, just a proprietary port channeling for switch interconnect as far as I can tell.
I suppose that leaves me with just using pure software (for now, I have an 8 port Intel Pro 100 at home that I'll look at...), but I'm unable to get the increased throughput using mode=0. I did notice, however, while I was on the switch console it kept complaining about the interfaces flapping and re-learning addresses. My network guy here at work said that it was bad and either the switch or the bond(s) is misconfigured somewhere. Any hints as to where and or what kinds of things I should be looking at?
--Tim
Hi!
LACP is a part of 802.3ad.
Cisco 2900XL supports 802.1Q, ISL, EtherChannel, LACP, etc. Depending on the age of the switch and the version and category of IOS used, the LACP support may be there. Cisco has many IOS versions for the hardware you have.
http://www.cisco.com/warp/public/620/1.html
As for protocols support, it's hard to beat Cisco.
You may have a look at:
http://www.cisco.com/warp/public/473/140.pdf http://www.cisco.com/en/US/products/hw/switches/ps607/products_configuration...
Hope this helped!
Guy Boisvert, ing. IngTegration inc.
On Fri, 2008-04-11 at 13:43 -0400, Guy Boisvert wrote:
Timothy Selivanow wrote:
On Thu, 2008-04-10 at 11:24 +0100, Michael Simpson wrote:
Certainly the 2900 will support 802.3ad or LACP natively.
i found this which may be of use
http://wiki.oracle.com/page/Cisco+Systems+IOS-based+switches-+interface+bonding+and+trunking?t=anon
So, as it turns out, it's a 2900XL, which does not support 802.3ad or LACP at all, just a proprietary port channeling for switch interconnect as far as I can tell.
I suppose that leaves me with just using pure software (for now, I have an 8 port Intel Pro 100 at home that I'll look at...), but I'm unable to get the increased throughput using mode=0. I did notice, however, while I was on the switch console it kept complaining about the interfaces flapping and re-learning addresses. My network guy here at work said that it was bad and either the switch or the bond(s) is misconfigured somewhere. Any hints as to where and or what kinds of things I should be looking at?
--Tim
Hi!
LACP is a part of 802.3ad.
Cisco 2900XL supports 802.1Q, ISL, EtherChannel, LACP, etc. Depending on the age of the switch and the version and category of IOS used, the LACP support may be there. Cisco has many IOS versions for the hardware you have.
And did you even try the command I give you? In the previous thread ---> Some of those are known to run IOS 10 - 12 The older 2900s don't support it from about 7 -8 years ago. Updates the IOS? If I paid a couple grand for a core switch i'd be calling Cisco!
http://wiki.oracle.com/page/Cisco+Systems+IOS-based+switches-+interface +bonding+and+trunking?t=anon> This seems like relevant info also....
http://www.cisco.com/warp/public/620/1.html
As for protocols support, it's hard to beat Cisco.
You may have a look at:
http://www.cisco.com/warp/public/473/140.pdf http://www.cisco.com/en/US/products/hw/switches/ps607/products_configuration...
Hope this helped!
Guy Boisvert, ing. IngTegration inc. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
--Tim ________________________________________________________ < God isn't dead, he just couldn't find a parking place. > -------------------------------------------------------- \ \ \ \ /\ ( ) .( o ).
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
They should run 12.x but you would have had to grab a copy before the support ran out. Maybe you have a newer image saved somewhere.
Les Mikesell wrote:
John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
They should run 12.x but you would have had to grab a copy before the support ran out. Maybe you have a newer image saved somewhere.
Cisco Internetwork Operating System Software IOS (tm) C2900XL Software (C2900XL-C3H2S-M), Version 12.0(5)WC10, RELEASE SOFTWARE (fc1) Copyright (c) 1986-2004 by cisco Systems, Inc. Compiled Fri 28-May-04 09:52 by antonino
I believe that was the last version for these puppies.
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
On Fri, 2008-04-11 at 15:14 -0400, John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
It's running IOS 12.0(5)WC16, but running that command in enable didn't work (also tried it in config). Thanks for all of your suggestions and help. It's still pretty frustrating though...
--Tim ____________________________ / "All Bibles are man-made." \ \ -- Thomas Edison / ---------------------------- \ \ \ \ /\ ( ) .( o ).
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 15:14 -0400, John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
It's running IOS 12.0(5)WC16, but running that command in enable didn't work (also tried it in config). Thanks for all of your suggestions and help. It's still pretty frustrating though...
sw1-ven-nyc(config-if)#switch trunk encap ? dot1q Interface uses only 802.1q trunking encapsulation when trunking isl Interface uses only ISL trunking encapsulation when trunking
It supports dot1q, you need to setup a port group for the 2 interfaces though and then configure trunking on that port group.
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
On Fri, 2008-04-11 at 15:46 -0400, Ross S. W. Walker wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 15:14 -0400, John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
It's running IOS 12.0(5)WC16, but running that command in enable didn't work (also tried it in config). Thanks for all of your suggestions and help. It's still pretty frustrating though...
sw1-ven-nyc(config-if)#switch trunk encap ? dot1q Interface uses only 802.1q trunking encapsulation when trunking isl Interface uses only ISL trunking encapsulation when trunking
It supports dot1q, you need to setup a port group for the 2 interfaces though and then configure trunking on that port group.
-Ross
It supports what you are after through using the "Expansion Modules" Like Ross said you will have to make create port Pairs IE, a 24 port switch can support up to 8 ports total . Maybe an easier way to accomplish this would be to use the cisco CMS or Cisco Works Manager.
Standards • IEEE 802.3x full duplex on 100Base-T ports • IEEE 802.1D Spanning-Tree • IEEE 802.3u 100Base-TX • IEEE 802.3 10Base-T specification • IEEE 802.3z 1000Base-X • IEEE 802.3ab • IEEE 802.1Q VLAN Trunking • ATM Forum LANE 1.0
Reference:
http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps607/product_dat...
This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
John wrote:
On Fri, 2008-04-11 at 15:46 -0400, Ross S. W. Walker wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 15:14 -0400, John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote: > If I paid a couple grand for a core switch i'd be calling Cisco! >
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
It's running IOS 12.0(5)WC16, but running that command in enable didn't work (also tried it in config). Thanks for all of your suggestions and help. It's still pretty frustrating though...
sw1-ven-nyc(config-if)#switch trunk encap ? dot1q Interface uses only 802.1q trunking encapsulation when trunking isl Interface uses only ISL trunking encapsulation when trunking
It supports dot1q, you need to setup a port group for the 2 interfaces though and then configure trunking on that port group.
-Ross
It supports what you are after through using the "Expansion Modules" Like Ross said you will have to make create port Pairs IE, a 24 port switch can support up to 8 ports total . Maybe an easier way to accomplish this would be to use the cisco CMS or Cisco Works Manager.
Standards
- IEEE 802.3x full duplex on 100Base-T ports
- IEEE 802.1D Spanning-Tree
- IEEE 802.3u 100Base-TX
- IEEE 802.3 10Base-T specification
- IEEE 802.3z 1000Base-X
- IEEE 802.3ab
- IEEE 802.1Q VLAN Trunking
- ATM Forum LANE 1.0
Reference:
http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps6 07/product_data_sheet09186a0080088a16.pdf
I don't think you need to do all that:
Check out: http://www.cisco.com/en/US/docs/switches/lan/catalyst2900xl_3500xl/release12...
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
Ross S. W. Walker wrote:
I don't think you need to do all that:
Check out: http://www.cisco.com/en/US/docs/switches/lan/catalyst2900xl_3500xl/release12...
I've run 2900xl's connected to each other though 2 ports each configured with: port group 1 distribution destination but never tried it with anything non-cisco.
Les Mikesell wrote:
Ross S. W. Walker wrote:
I don't think you need to do all that:
Check out:
http://www.cisco.com/en/US/docs/switches/lan/catalyst2900xl_3500xl/release12...
I've run 2900xl's connected to each other though 2 ports each configured with: port group 1 distribution destination but never tried it with anything non-cisco.
This should give you what you need when doing 802.3ad LAGs:
interface FastEthernet0/21 port group 1 spanning-tree portfast ! interface FastEthernet0/22 port group 1 spanning-tree portfast !
It might have been the lack of STP portfasting that gave you trouble initially as the Linux LACP might have timed out waiting for STP discovery to finish.
Here is a good wiki article on link aggregation: http://en.wikipedia.org/wiki/Link_aggregation
And here is a good site on Linux bonding interface: http://www.linux-foundation.org/en/Net:Bonding
You only need to setup port groups when doing 802.3ad, if you use the other methods you may need to tune out the 2900XL for the constantly shifting MAC addresses.
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
On Fri, 2008-04-11 at 17:23 -0400, Ross S. W. Walker wrote:
Les Mikesell wrote:
Ross S. W. Walker wrote:
I don't think you need to do all that:
Check out:
http://www.cisco.com/en/US/docs/switches/lan/catalyst2900xl_3500xl/release12...
I've run 2900xl's connected to each other though 2 ports each configured with: port group 1 distribution destination but never tried it with anything non-cisco.
This should give you what you need when doing 802.3ad LAGs:
interface FastEthernet0/21 port group 1 spanning-tree portfast ! interface FastEthernet0/22 port group 1 spanning-tree portfast !
It might have been the lack of STP portfasting that gave you trouble initially as the Linux LACP might have timed out waiting for STP discovery to finish.
Here is a good wiki article on link aggregation: http://en.wikipedia.org/wiki/Link_aggregation
I did not know linux used a higher part of the networking stack to negotiate protocols. That's interesting.
And here is a good site on Linux bonding interface: http://www.linux-foundation.org/en/Net:Bonding
You only need to setup port groups when doing 802.3ad, if you use the other methods you may need to tune out the 2900XL for the constantly shifting MAC addresses.
-Ross
This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Fri, 2008-04-11 at 17:23 -0400, Ross S. W. Walker wrote:
This should give you what you need when doing 802.3ad LAGs:
interface FastEthernet0/21 port group 1 spanning-tree portfast ! interface FastEthernet0/22 port group 1 spanning-tree portfast !
Using this on the switch and mode=4 for bonding results in the machine not being able to contact anything (happens with just 'port group 1' too, I tried that a "long" time ago), but there aren't any errors. So, it doesn't work, but I couldn't tell you why. Is there anything else that needs to be done, perhaps on the host's side?
--Tim _________________________________________________________________________ / You know it's going to be a bad day when you want to put on the clothes \ \ you wore home from the party and there aren't any. / ------------------------------------------------------------------------- \ \ \ \ /\ ( ) .( o ).
On Fri, 2008-04-11 at 17:13 -0700, Timothy Selivanow wrote:
On Fri, 2008-04-11 at 17:23 -0400, Ross S. W. Walker wrote:
This should give you what you need when doing 802.3ad LAGs:
interface FastEthernet0/21 port group 1 spanning-tree portfast ! interface FastEthernet0/22 port group 1 spanning-tree portfast !
Using this on the switch and mode=4 for bonding results in the machine not being able to contact anything (happens with just 'port group 1' too, I tried that a "long" time ago), but there aren't any errors. So, it doesn't work, but I couldn't tell you why. Is there anything else that needs to be done, perhaps on the host's side?
Take a look at: http://wiki.centos.org/TipsAndTricks/BondingInterfaces?highlight=% 28Bonding%29
1. Can you try at least try the other Bonding Modes please? 2. Try another Switch? 3. What info I can glean is the 2dot3ad will work on the IOS version you have. 4. Did you validate that Both NIC Cards in the Server works and there not like flaking out? 5. Another thought is are they the same brand? Linux it should not matter but you never know. 6. Validate the NICS working separatly? Instead of in Bonding mode? 7. Faulty cableling, connectors and such. 8. Do you have the bunding configured right on the server?
Some of questions may sound rhetorical and stupid but I have found many times it was the simplist thing that was wrong. The link on the wiki bonding will work. It is a good example, clear and concise.
--Tim _________________________________________________________________________ / You know it's going to be a bad day when you want to put on the clothes \ \ you wore home from the party and there aren't any. /
\ \ \ \ /\ ( ) .( o ).
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 17:23 -0400, Ross S. W. Walker wrote:
This should give you what you need when doing 802.3ad LAGs:
interface FastEthernet0/21 port group 1 spanning-tree portfast ! interface FastEthernet0/22 port group 1 spanning-tree portfast !
Using this on the switch and mode=4 for bonding results in the machine not being able to contact anything (happens with just 'port group 1' too, I tried that a "long" time ago), but there aren't any errors. So, it doesn't work, but I couldn't tell you why. Is there anything else that needs to be done, perhaps on the host's side?
Does the same thing happen if you only connect one interface? I was under the impression that a single member of a group would work the same way as a normal interface. Also does it help to add 'distribution destination'?
On Fri, 2008-04-11 at 19:45 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 17:23 -0400, Ross S. W. Walker wrote:
This should give you what you need when doing 802.3ad LAGs:
interface FastEthernet0/21 port group 1 spanning-tree portfast ! interface FastEthernet0/22 port group 1 spanning-tree portfast !
Using this on the switch and mode=4 for bonding results in the machine not being able to contact anything (happens with just 'port group 1' too, I tried that a "long" time ago), but there aren't any errors. So, it doesn't work, but I couldn't tell you why. Is there anything else that needs to be done, perhaps on the host's side?
Does the same thing happen if you only connect one interface? I was under the impression that a single member of a group would work the same way as a normal interface. Also does it help to add 'distribution destination'?
Have you tried Port Group 2 - 8. Hardware Failing? Those are the right commands for it as I found in the cisco docs.
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 17:23 -0400, Ross S. W. Walker wrote:
This should give you what you need when doing 802.3ad LAGs:
interface FastEthernet0/21 port group 1 spanning-tree portfast ! interface FastEthernet0/22 port group 1 spanning-tree portfast !
Using this on the switch and mode=4 for bonding results in the machine not being able to contact anything (happens with just 'port group 1' too, I tried that a "long" time ago), but there aren't any errors. So, it doesn't work, but I couldn't tell you why. Is there anything else that needs to be done, perhaps on the host's side?
Can you post the ifcfg files used and the output of /proc/net/bonding/bond0?
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
On Sat, 2008-04-12 at 11:06 -0400, Ross S. W. Walker wrote:
Can you post the ifcfg files used and the output of /proc/net/bonding/bond0?
This is for one system. I have another one that I've been working on too, and it too doesn't work with 'port group 2' on the two switch ports that it is connected to (haven't tried adding 'spanning-tree portfast' to those yet). I've been focusing on this one first as it is an iSCSI target, and the other one is running Xen so it's a bit more complicated (DomU can't get out at the moment with the bond, but Dom0 can, still no increased throughput...)
--Tim _________________________________________________________________ / A political man can have as his aim the realization of freedom, \ | but he has no means to realize it other than through violence. | \ -- Jean Paul Sartre / ----------------------------------------------------------------- \ \ \ \ /\ ( ) .( o ).
On Mon, 2008-04-14 at 10:48 -0700, Timothy Selivanow wrote:
On Sat, 2008-04-12 at 11:06 -0400, Ross S. W. Walker wrote:
Can you post the ifcfg files used and the output of /proc/net/bonding/bond0?
This is for one system. I have another one that I've been working on too, and it too doesn't work with 'port group 2' on the two switch ports that it is connected to (haven't tried adding 'spanning-tree portfast' to those yet). I've been focusing on this one first as it is an iSCSI target, and the other one is running Xen so it's a bit more complicated (DomU can't get out at the moment with the bond, but Dom0 can, still no increased throughput...)
I've changed the switch out, unfortunately to something that I know doesn't support 802.3ad, but I'm still unable to get aggregate link bandwidth using mode 0, 2, and 6. I'm using scp to test the bandwidth, one machine with one interface, one with two bonded, and one with three bonded. No matter the combination of who is sending/receiving the files, no increase in throughput.
Would using a x-over cable on two machines, using two interfaces each, with 802.3ad (or other mode...) on both hosts work? My inclination is that the aggregating protocol needs a shared bus to negotiate, and putting each channel on it's own bus (x-over cable) would defeat that...
--Tim ______________________________________________ < Support Bingo, keep Grandma off the streets. > ---------------------------------------------- \ \ \ \ /\ ( ) .( o ).
Timothy Selivanow wrote:
I've changed the switch out, unfortunately to something that I know doesn't support 802.3ad, but I'm still unable to get aggregate link bandwidth using mode 0, 2, and 6. I'm using scp to test the bandwidth, one machine with one interface, one with two bonded, and one with three bonded. No matter the combination of who is sending/receiving the files, no increase in throughput.
Would using a x-over cable on two machines, using two interfaces each, with 802.3ad (or other mode...) on both hosts work? My inclination is that the aggregating protocol needs a shared bus to negotiate, and putting each channel on it's own bus (x-over cable) would defeat that...
AFAIK, bonding can not give increased bandwidth between two hosts - the maximum you can ever get is the bandwidth of one of the links i.e. if you have a server with say 4 bonded interfaces, any one client can only get a maximum bandwidth of one of the interfaces on the server.
I've used 2 bonded (mode 6) Gigabit interfaces on NFS servers and can get 200+Mbyte/s read speeds using multiple clients
James Pearson
Just out of curiosity.
If you wanted to bond do you have to ask your network admin to configure a special switch setting for MAC addresses?
On Tue, Apr 15, 2008 at 6:27 PM, James Pearson james-p@moving-picture.com wrote:
Timothy Selivanow wrote:
I've changed the switch out, unfortunately to something that I know doesn't support 802.3ad, but I'm still unable to get aggregate link bandwidth using mode 0, 2, and 6. I'm using scp to test the bandwidth, one machine with one interface, one with two bonded, and one with three bonded. No matter the combination of who is sending/receiving the files, no increase in throughput.
Would using a x-over cable on two machines, using two interfaces each, with 802.3ad (or other mode...) on both hosts work? My inclination is that the aggregating protocol needs a shared bus to negotiate, and putting each channel on it's own bus (x-over cable) would defeat that...
AFAIK, bonding can not give increased bandwidth between two hosts - the maximum you can ever get is the bandwidth of one of the links i.e. if you have a server with say 4 bonded interfaces, any one client can only get a maximum bandwidth of one of the interfaces on the server.
I've used 2 bonded (mode 6) Gigabit interfaces on NFS servers and can get 200+Mbyte/s read speeds using multiple clients
James Pearson
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On 5/22/08, Mag Gam magawake@gmail.com wrote:
Just out of curiosity.
If you wanted to bond do you have to ask your network admin to configure a special switch setting for MAC addresses?
depends on the mode of bonding
Thanks Jim. Since, 802.3ad requires switch settings does it perform better than other modes? Does anyone have any benchmarks?
TIA
On Thu, May 22, 2008 at 7:19 AM, James Pearson james-p@moving-picture.com wrote:
Mag Gam wrote:
Just out of curiosity.
If you wanted to bond do you have to ask your network admin to configure a special switch setting for MAC addresses?
AFAIK, only with 802.3ad
The other Linux bonding modes don't require any switch settings
James Pearson
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Mag Gam wrote:
Thanks Jim. Since, 802.3ad requires switch settings does it perform better than other modes? Does anyone have any benchmarks?
I haven't done any benchmarks - but as I've managed to get 200+Mbyte/s read speeds using mode 6 with a dual link - I can't see it would be any faster with 802.3ad
James Pearson
On Fri, May 23, 2008 at 6:23 AM, James Pearson james-p@moving-picture.com wrote:
Mag Gam wrote:
Thanks Jim. Since, 802.3ad requires switch settings does it perform better than other modes? Does anyone have any benchmarks?
I haven't done any benchmarks - but as I've managed to get 200+Mbyte/s read speeds using mode 6 with a dual link - I can't see it would be any faster with 802.3ad
Mode 6 (AFAIR) requires you to connect both cables to the same switch. 802.3ad allows you to get link aggregation while connecting to different switches, which means you're still up if one of the switches dies (providing you connect all the hosts to both switches). The switches themselves need to be interconnected (in 3COM switches, with a proprietary cable) to be able to span 802.3ad links between switches, but it's a very useful feature anyway.
Filipe
On Fri, 2008-04-11 at 16:59 -0400, Ross S. W. Walker wrote:
John wrote:
On Fri, 2008-04-11 at 15:46 -0400, Ross S. W. Walker wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 15:14 -0400, John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote: > On Fri, 2008-04-11 at 14:26 -0400, John wrote: >> If I paid a couple grand for a core switch i'd be calling Cisco! >> > > They're not core switches, they're just the catalyst 2900XL which is a > small 24 port switch. We've got stacks of these sitting unused, so I > doubt they're too expensive/valuable. >
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
It's running IOS 12.0(5)WC16, but running that command in enable didn't work (also tried it in config). Thanks for all of your suggestions and help. It's still pretty frustrating though...
sw1-ven-nyc(config-if)#switch trunk encap ? dot1q Interface uses only 802.1q trunking encapsulation when trunking isl Interface uses only ISL trunking encapsulation when trunking
It supports dot1q, you need to setup a port group for the 2 interfaces though and then configure trunking on that port group.
-Ross
It supports what you are after through using the "Expansion Modules" Like Ross said you will have to make create port Pairs IE, a 24 port switch can support up to 8 ports total . Maybe an easier way to accomplish this would be to use the cisco CMS or Cisco Works Manager.
Standards
- IEEE 802.3x full duplex on 100Base-T ports
- IEEE 802.1D Spanning-Tree
- IEEE 802.3u 100Base-TX
- IEEE 802.3 10Base-T specification
- IEEE 802.3z 1000Base-X
- IEEE 802.3ab
- IEEE 802.1Q VLAN Trunking
- ATM Forum LANE 1.0
Reference:
http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps6 07/product_data_sheet09186a0080088a16.pdf
I don't think you need to do all that:
Check out: http://www.cisco.com/en/US/docs/switches/lan/catalyst2900xl_3500xl/release12...
-Ross
This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
I Qoute: "Trunking is supported on all 8-MB switches running Cisco IOS 12.0(5)XP and later. Trunking is not supported on some older software releases and on some older Catalyst 2900 XL switches and modules. For information about which older devices and software releases support trunking, refer to the release notes for Release 11.2(8)SA6 or earlier (http://www.cisco.com/univercd/cc/td/doc/product/lan/c2900xl/index.htm)."
References: http://www.cisco.com/en/US/docs/switches/lan/catalyst2900xl_3500xl/release12...
Now for what version of the 2900 you have I do not know. But it seems the info on Ciscos site is kind of misleading in places. There is some documentation that says it works and some say you have to have the add on modules.
Your best bet if your not comfortable with the IOS command line is to use the Web Interface or the Cisco Works Manager Interface. I say this because I have really no way of knowing your experiance on this. I wouln't do this on a live production switch.
If any one else could chime in and provide a 2900 configuration with trunking setup would be nice. Just remove your IPs and replace with x.x.x.x. I am interested in this also.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
John wrote:
Now for what version of the 2900 you have I do not know. But it seems the info on Ciscos site is kind of misleading in places. There is some documentation that says it works and some say you have to have the add on modules.
Your best bet if your not comfortable with the IOS command line is to use the Web Interface or the Cisco Works Manager Interface. I say this because I have really no way of knowing your experiance on this. I wouln't do this on a live production switch.
Trunking has to do with carrying multiple tagged vlans on one port - and the 2900xl with a 12.x IOS should do that in either ISL (cisco) or dot1q mode. But I thought we were grouping ports instead. You should be able to do both, but the interfaces in the same port group would have to have the same trunk encapsulation set. There is a separate 'vlan database' command on those switches where you have to add the vlan numbers that you want a trunk to carry - and it doesn't get saved in the visible config file. If you have 'switchport access vlan nnn' on any interface that number goes in the vlan database automatically but you just want the trunk ports to pass some other vlans through you have to add them explicitly.
On Fri, 2008-04-11 at 18:05 -0500, Les Mikesell wrote:
John wrote:
Now for what version of the 2900 you have I do not know. But it seems the info on Ciscos site is kind of misleading in places. There is some documentation that says it works and some say you have to have the add on modules.
Your best bet if your not comfortable with the IOS command line is to use the Web Interface or the Cisco Works Manager Interface. I say this because I have really no way of knowing your experiance on this. I wouln't do this on a live production switch.
Trunking has to do with carrying multiple tagged vlans on one port - and the 2900xl with a 12.x IOS should do that in either ISL (cisco) or dot1q mode. But I thought we were grouping ports instead. You should be able to do both, but the interfaces in the same port group would have to have the same trunk encapsulation set. There is a separate 'vlan database' command on those switches where you have to add the vlan numbers that you want a trunk to carry - and it doesn't get saved in the visible config file. If you have 'switchport access vlan nnn' on any interface that number goes in the vlan database automatically but you just want the trunk ports to pass some other vlans through you have to add them explicitly.
' Ok, if I am reading your reply right VLAN configs are so to say saved in the switch instead of the config file correct? So a config file never contains the VLAN info. Just wondering any way to copy or backup the VLAN Database? There has to be shouldn't. Been 4 years or more since I've touched one. Guess I need to get the newest emulator software Cisco has.
John wrote:
Now for what version of the 2900 you have I do not know. But it seems the info on Ciscos site is kind of misleading in places. There is some documentation that says it works and some say you have to have the add on modules.
Your best bet if your not comfortable with the IOS command line is to use the Web Interface or the Cisco Works Manager Interface. I say this because I have really no way of knowing your experiance on this. I wouln't do this on a live production switch.
Trunking has to do with carrying multiple tagged vlans on one port - and the 2900xl with a 12.x IOS should do that in either ISL (cisco) or dot1q mode. But I thought we were grouping ports instead. You should be able to do both, but the interfaces in the same port group would have to have the same trunk encapsulation set. There is a separate 'vlan database' command on those switches where you have to add the vlan numbers that you want a trunk to carry - and it doesn't get saved in the visible config file. If you have 'switchport access vlan nnn' on any interface that number goes in the vlan database automatically but you just want the trunk ports to pass some other vlans through you have to add them explicitly.
' Ok, if I am reading your reply right VLAN configs are so to say saved in the switch instead of the config file correct? So a config file never contains the VLAN info. Just wondering any way to copy or backup the VLAN Database? There has to be shouldn't. Been 4 years or more since I've touched one. Guess I need to get the newest emulator software Cisco has.
Actually it is more complicated that that - the switches have a vtp master/slave concept to learn vlans but I've always turned it off because the switch with the latest change becomes the master so if someone swaps a switch from the lab into production it would probably break everything in sight. As far as backing it up goes, I just keep a text file that goes: vlan database vlan nnn vlan nnn etc. exit to paste into a new or moved switch. But I think the new models do something different.
On Fri, 2008-04-11 at 19:51 -0500, Les Mikesell wrote:
John wrote:
Now for what version of the 2900 you have I do not know. But it seems the info on Ciscos site is kind of misleading in places. There is some documentation that says it works and some say you have to have the add on modules.
Your best bet if your not comfortable with the IOS command line is to use the Web Interface or the Cisco Works Manager Interface. I say this because I have really no way of knowing your experiance on this. I wouln't do this on a live production switch.
Trunking has to do with carrying multiple tagged vlans on one port - and the 2900xl with a 12.x IOS should do that in either ISL (cisco) or dot1q mode. But I thought we were grouping ports instead. You should be able to do both, but the interfaces in the same port group would have to have the same trunk encapsulation set. There is a separate 'vlan database' command on those switches where you have to add the vlan numbers that you want a trunk to carry - and it doesn't get saved in the visible config file. If you have 'switchport access vlan nnn' on any interface that number goes in the vlan database automatically but you just want the trunk ports to pass some other vlans through you have to add them explicitly.
' Ok, if I am reading your reply right VLAN configs are so to say saved in the switch instead of the config file correct? So a config file never contains the VLAN info. Just wondering any way to copy or backup the VLAN Database? There has to be shouldn't. Been 4 years or more since I've touched one. Guess I need to get the newest emulator software Cisco has.
Actually it is more complicated that that - the switches have a vtp master/slave concept to learn vlans but I've always turned it off because the switch with the latest change becomes the master so if someone swaps a switch from the lab into production it would probably break everything in sight. As far as backing it up goes, I just keep a text file that goes: vlan database vlan nnn vlan nnn etc. exit to paste into a new or moved switch. But I think the new models do something different.
So he could take a preconfigured configuration of the port config and the vlan and copy it to the VLAN Database to use it. That is the way I always done the configuration if I had acess to a tftp server. Of course to do that now for as command wise I've no clue for the vlan database/
On Fri, 2008-04-11 at 12:33 -0700, Timothy Selivanow wrote:
On Fri, 2008-04-11 at 15:14 -0400, John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
It's running IOS 12.0(5)WC16, but running that command in enable didn't work (also tried it in config). Thanks for all of your suggestions and help. It's still pretty frustrating though...
You mean ]#configt
--Tim ____________________________ / "All Bibles are man-made." \ \ -- Thomas Edison /
\ \ \ \ /\ ( ) .( o ).
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Fri, 2008-04-11 at 12:33 -0700, Timothy Selivanow wrote:
On Fri, 2008-04-11 at 15:14 -0400, John wrote:
On Fri, 2008-04-11 at 14:08 -0500, Les Mikesell wrote:
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
They were expensive once. Now they are end-of-life'd and you can't get a support contract or a new IOS for them.
Yea there 8years old and do not support todays technology. All the way back to IOS 8..
It's running IOS 12.0(5)WC16, but running that command in enable didn't work (also tried it in config). Thanks for all of your suggestions and help. It's still pretty frustrating though...
The version of IOS you have running supports that command. But that is hardware dependent also.
Just out of being currious you can export the configuration to tftp, add the appropriate section into the config file, tehn copy it back to the switch. It will either work or NOT!!! That's is up to you. Since you all have a bunch laying around.
--Tim ____________________________ / "All Bibles are man-made." \ \ -- Thomas Edison /
\ \ \ \ /\ ( ) .( o ).
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Fri, 2008-04-11 at 11:49 -0700, Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
From what I was taught all Catalyst Switched and Routers are of "Core
Type" Some just not the fancy functionality you desire.
If you look at my previous post I listed what was supported by the 2900 switch. Then I give you a command to ""try"". And I didn't say the command would work either. If it worked you would have knew off the bat that the switch supported it.
--Tim ________________________________________________________
< God isn't dead, he just couldn't find a parking place. >
\ \ \ \ /\ ( ) .( o ).
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Timothy Selivanow wrote:
On Fri, 2008-04-11 at 14:26 -0400, John wrote:
If I paid a couple grand for a core switch i'd be calling Cisco!
They're not core switches, they're just the catalyst 2900XL which is a small 24 port switch. We've got stacks of these sitting unused, so I doubt they're too expensive/valuable.
I use those for vendor inter-connects. Basically just provide VLANs for vendor applications we direct connect to.
-Ross
______________________________________________________________________ This e-mail, and any attachments thereto, is intended only for use by the addressee(s) named herein and may contain legally privileged and/or confidential information. If you are not the intended recipient of this e-mail, you are hereby notified that any dissemination, distribution or copying of this e-mail, and any attachments thereto, is strictly prohibited. If you have received this e-mail in error, please immediately notify the sender and permanently delete the original and any copy or printout thereof.
On Fri, 2008-04-11 at 13:43 -0400, Guy Boisvert wrote:
You may have a look at:
http://www.cisco.com/warp/public/473/140.pdf http://www.cisco.com/en/US/products/hw/switches/ps607/products_configuration...
I've looked at both of these documents already. In the PDF, on page 3, it says that the 2900XL does not support LACP nor PAgP (any IOS version). In all of the configuration examples that I've seen (none of them use a 2900XL in the example, or I'm dense which is completely plausible), none of that specific syntax applies (command unknown, etc...). The switch is using an old version of IOS, so I'll look into an upgrade path, but I'd rather move onto my project than mess with cisco stuff ;)
--Tim _________________________________________________________________________ / The 11 is for people with the pride of a 10 and the pocketbook of an 8. \ \ -- R.B. Greenberg [referring to PDPs?] / ------------------------------------------------------------------------- \ \ \ \ /\ ( ) .( o ).
-
Filipe Brandenburger -
Guy Boisvert -
James Pearson -
John -
Joseph L. Casale -
Les Mikesell -
Mag Gam -
Michael Simpson -
Ross S. W. Walker -
Timothy Selivanow