Does anyone here on this list have experience with HA clustering?
I'm previewing drbd as a potential tool, and wanted to know if anyone here has experiemented with it at all... How stable is it? Does the additional likelyhood of failure given the additional complexity actually get compensated by a better overall system?
Any feedback is welcome...
-Ben
On 10/03/06, Benjamin Smith lists@benjamindsmith.com wrote:
Does anyone here on this list have experience with HA clustering?
I'm previewing drbd as a potential tool, and wanted to know if anyone here has experiemented with it at all... How stable is it? Does the additional likelyhood of failure given the additional complexity actually get compensated by a better overall system?
I've used DRBD and Heartbeat in various guises for various roles over the last 5 or 6 years. Initially I was loathe to put it in production, it just didn't seem polished enough.
Nowadays it's pretty decent though we don't use it for vast quantities of data.
We have some old-ish boxes running as LVS loadbalancers for a handful of mail and webservers and these are pretty solid.
We run DRBD/Heartbeat clusters for a Qmail/VPOPMail NFS mailstore which holds around 40GB of customer Maildirs. That's running the packaged DRBD and Heartbeat RPMs from CentOS Extras and has been solid since we switched back to NFS3 from NFS4.
We run a similar setup with small MySQL and Postgres databases and that's pretty reliable too.
Will.
Ugh. What a week!
Anyway, my situation is that we have a production server in San Fransisco, and a "hot" backup in my hometown (Chico, CA) .
What I'd like to do is mirror the production server to the local one, so that if the SF server goes down, we have work saved to the last possible moment. Say, within 10 minutes.... Is this feasible?
Thanks,
=Ben
On Friday 10 March 2006 03:43, Will McDonald wrote:
On 10/03/06, Benjamin Smith lists@benjamindsmith.com wrote:
Does anyone here on this list have experience with HA clustering?
I'm previewing drbd as a potential tool, and wanted to know if anyone here
has
experiemented with it at all... How stable is it? Does the additional likelyhood of failure given the additional complexity actually get compensated by a better overall system?
I've used DRBD and Heartbeat in various guises for various roles over the last 5 or 6 years. Initially I was loathe to put it in production, it just didn't seem polished enough.
Nowadays it's pretty decent though we don't use it for vast quantities of
data.
We have some old-ish boxes running as LVS loadbalancers for a handful of mail and webservers and these are pretty solid.
We run DRBD/Heartbeat clusters for a Qmail/VPOPMail NFS mailstore which holds around 40GB of customer Maildirs. That's running the packaged DRBD and Heartbeat RPMs from CentOS Extras and has been solid since we switched back to NFS3 from NFS4.
We run a similar setup with small MySQL and Postgres databases and that's pretty reliable too.
Will. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
On Thu, 2006-03-16 at 23:52 -0800, Benjamin Smith wrote:
Ugh. What a week!
Anyway, my situation is that we have a production server in San Fransisco, and a "hot" backup in my hometown (Chico, CA) .
What I'd like to do is mirror the production server to the local one, so that if the SF server goes down, we have work saved to the last possible moment. Say, within 10 minutes.... Is this feasible?
I don't think something like DRBD is going to work very well across a WAN link. The amount of traffic generated by drbd can be pretty large, it is enough that I normally use a gigabit crossover cable between 2 servers (if possible) when using drbd on them.
I would think that rsyncs of the appropriate directories at a period in time might be the best way to handle this.
I am getting ready to do this in then next week or so myself ... if I have any luck, I'll tell you what solution I found. In my case I am also worried about a mysql database that has live info in it ... and an ldap database too.
On Friday 10 March 2006 03:43, Will McDonald wrote:
On 10/03/06, Benjamin Smith lists@benjamindsmith.com wrote:
Does anyone here on this list have experience with HA clustering?
I'm previewing drbd as a potential tool, and wanted to know if anyone here
has
experiemented with it at all... How stable is it? Does the additional likelyhood of failure given the additional complexity actually get compensated by a better overall system?
I've used DRBD and Heartbeat in various guises for various roles over the last 5 or 6 years. Initially I was loathe to put it in production, it just didn't seem polished enough.
Nowadays it's pretty decent though we don't use it for vast quantities of
data.
We have some old-ish boxes running as LVS loadbalancers for a handful of mail and webservers and these are pretty solid.
We run DRBD/Heartbeat clusters for a Qmail/VPOPMail NFS mailstore which holds around 40GB of customer Maildirs. That's running the packaged DRBD and Heartbeat RPMs from CentOS Extras and has been solid since we switched back to NFS3 from NFS4.
We run a similar setup with small MySQL and Postgres databases and that's pretty reliable too.
Will. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
Is there any plan to include heartbeat 2.0.4 rpm and drbd last version in extras?
On 3/17/06, Johnny Hughes mailing-lists@hughesjr.com wrote:
On Thu, 2006-03-16 at 23:52 -0800, Benjamin Smith wrote:
Ugh. What a week!
Anyway, my situation is that we have a production server in San
Fransisco, and
a "hot" backup in my hometown (Chico, CA) .
What I'd like to do is mirror the production server to the local one, so
that
if the SF server goes down, we have work saved to the last possible
moment.
Say, within 10 minutes.... Is this feasible?
I don't think something like DRBD is going to work very well across a WAN link. The amount of traffic generated by drbd can be pretty large, it is enough that I normally use a gigabit crossover cable between 2 servers (if possible) when using drbd on them.
I would think that rsyncs of the appropriate directories at a period in time might be the best way to handle this.
I am getting ready to do this in then next week or so myself ... if I have any luck, I'll tell you what solution I found. In my case I am also worried about a mysql database that has live info in it ... and an ldap database too.
On Friday 10 March 2006 03:43, Will McDonald wrote:
On 10/03/06, Benjamin Smith lists@benjamindsmith.com wrote:
Does anyone here on this list have experience with HA clustering?
I'm previewing drbd as a potential tool, and wanted to know if
anyone here
has
experiemented with it at all... How stable is it? Does the
additional
likelyhood of failure given the additional complexity actually get compensated by a better overall system?
I've used DRBD and Heartbeat in various guises for various roles over the last 5 or 6 years. Initially I was loathe to put it in production, it just didn't seem polished enough.
Nowadays it's pretty decent though we don't use it for vast quantities
of
data.
We have some old-ish boxes running as LVS loadbalancers for a handful of mail and webservers and these are pretty solid.
We run DRBD/Heartbeat clusters for a Qmail/VPOPMail NFS mailstore which holds around 40GB of customer Maildirs. That's running the packaged DRBD and Heartbeat RPMs from CentOS Extras and has been solid since we switched back to NFS3 from NFS4.
We run a similar setup with small MySQL and Postgres databases and that's pretty reliable too.
Will. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.6 (GNU/Linux)
iD8DBQBEGnO8TKkMgmrBY7MRArazAKChc8VMo1/jtTS2HkaI7FcrARL6CACgjWR1 4e5xMghZfSnMt/m90fgl8D0= =bEH0 -----END PGP SIGNATURE-----
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Fri, 2006-03-17 at 10:24 +0100, Alberto wrote:
Is there any plan to include heartbeat 2.0.4 rpm and drbd last version in extras?
It is not currently included latest stable version of Ultra Monkey 3, which is where the heartbeat part of drbd for centos extras comes from.
http://www.ultramonkey.org/download/3/
Is there something that you specifically need heartbeat 2 for that is not provided in the stable UM3 release?
That release includes heartbeat 1.2.3 . I want to use heartbeat 2 that is a complete rewrite with lots of new features, like resource monitoring.
On 3/17/06, Johnny Hughes mailing-lists@hughesjr.com wrote:
On Fri, 2006-03-17 at 10:24 +0100, Alberto wrote:
Is there any plan to include heartbeat 2.0.4 rpm and drbd last version in extras?
It is not currently included latest stable version of Ultra Monkey 3, which is where the heartbeat part of drbd for centos extras comes from.
http://www.ultramonkey.org/download/3/
Is there something that you specifically need heartbeat 2 for that is not provided in the stable UM3 release?
-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.6 (GNU/Linux)
iD8DBQBEGoyyTKkMgmrBY7MRAje+AJsH/xLh4g0kMhLm5ApcTwukDYWq4wCgq/aG rUqaw5j85d/S4ENtCBBc6pE= =juf5 -----END PGP SIGNATURE-----
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-----Original Message----- From: centos-bounces@centos.org [mailto:centos-bounces@centos.org]On Behalf Of Johnny Hughes Sent: Friday, March 17, 2006 3:31 AM To: CentOS ML Subject: Re: [CentOS] High-Availability Clustering and drbd?
On Thu, 2006-03-16 at 23:52 -0800, Benjamin Smith wrote:
Ugh. What a week!
Anyway, my situation is that we have a production server in San
Fransisco, and
a "hot" backup in my hometown (Chico, CA) .
What I'd like to do is mirror the production server to the
local one, so that
if the SF server goes down, we have work saved to the last
possible moment.
Say, within 10 minutes.... Is this feasible?
I don't think something like DRBD is going to work very well across a WAN link. The amount of traffic generated by drbd can be pretty large, it is enough that I normally use a gigabit crossover cable between 2 servers (if possible) when using drbd on them.
I would think that rsyncs of the appropriate directories at a period in time might be the best way to handle this.
I am getting ready to do this in then next week or so myself ... if I have any luck, I'll tell you what solution I found. In my case I am also worried about a mysql database that has live info in it ... and an ldap database too.
Good to know the bandwidth requirements of DRBD before trying it over our WAN. We use rsync to keep a directory structure in sync with a backup server on the other side of the contry. In total the directory is about 34GB and consists of images of scanned documents ov verious sizes. We shoot this over a T1 across the US and once the original sync is done, it does a nice job a couple times a day checking and updating differences.
I have also played with Unison (http://www.cis.upenn.edu/~bcpierce/unison/) with some success, but ended up using rsync for another project.
As for MySQL over WAN, how have people done with the built in replication?
Andrew
-
Alberto -
Andrew Cotter -
Benjamin Smith -
Johnny Hughes -
Will McDonald