Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
- CS.
------------------- # For EL5 virtual machines, Append the following in Grub to help keep the clock from drifting # and to reduce the interupt requests # 32bit: --append="rhgb quiet divider=10 clocksource=acpi_pm" # 64bit: --append="rhgb quiet notsc divider=10" bootloader --location=mbr --md5pass=$1$mXSD1l6mO$BBCk1gYArAATS7dlCQGthN. --append="rhgb quiet divider=10 clocksource=acpi_pm"
%packages --nobase # Other packages not listed here # ntp was installed ntp
### Add step-tickers ### cat > /etc/ntp/step-tickers <<\EOF2 0.centos.pool.ntp.org 1.centos.pool.ntp.org EOF2 ### End of step-ticker file ### --------------------
# Applied patch mentioned on CentOS page # http://wiki.centos.org/TipsAndTricks/VMWare_Server patch --verbose -b -l -i /root/ntp.patch
--------------------
VMware Tools not installed.
--------------------
Thanks, CS.
Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
CS.
I'm not sure what version of VMware Server you're running, but I had issues keeping accurate time within guest VMs until I followed the instructions at:
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd...
If your guest clocks are running too quickly, this may apply to you.
-Greg
On Tue, Oct 13, 2009 at 2:42 PM, Carlos Santana neubyr@gmail.com wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
The knowledgebase article referenced in the VMWare Server wiki article:
http://kb.vmware.com/kb/1006427
is the best source as far as I can tell. This is because VMWare keep the contents up-to-date. As suggested in there, you may want to try the 5.4 kernel -164 (already available for CentOS) without using clocksource or divider= options. This kernel has patches that were offered by VMWare developers to address the time drift issues.
Akemi
Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
[..]
VMware Tools not installed.
You should certainly install vmware tools, and enable time sync to the guest. Also don't run an ntp server in a Vmware VM.
I had an issue on my system running Debian and VMware server, I think it was a hardware/bios issue but the thing was the host system detected the clock speed of the hardware at 1/2 the proper speed. Which caused the host OS clock speed to double. Which wrecked havok in the VMs. Once I loaded the 'p4-clockmod' module that 'fixed' the clock speed in the host and I restarted the guest VMs and things were good after that.
Another thing to check is to make sure the 'rtc' driver is loaded, my recent vmware server experience is limited to running it on Debian(have it on 2 systems), so I can't speak to running it on top of CentOS. Most of my ESX guests are CentOS though.
Another thing I do is have ntpdate run in cron every so often to do another 'sync' to the host, typically once every 5 minutes I found that in my experience at least there is still some drift over time without doing that even with vmware tools time sync enabled. I do the same on guests running on ESX(roughly 300 of them).
I have found on ESX at least, haven't tried any other version of vmware, but on ESX with a VMI enabled kernel(unfortunately none of the RHEL4/5 kernels are VMI-enabled) with paravirtualization you can run an ntp server in the guest. I run dedicated Fedora 8 VMs with ntp servers(my vmware servers sync against those VMs), for this purpose. I know paravirtualization is going away in VMware at some point, hoping to find another solution before that happens.
nate
On Tue, Oct 13, 2009 at 04:31:03PM -0700, nate wrote:
Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
[..]
VMware Tools not installed.
You should certainly install vmware tools, and enable time sync to the guest. Also don't run an ntp server in a Vmware VM.
This is what I'd always thought, but the VMware KB link[1] referenced in the other reply in this thread seems to indicate that best practice is to use NTP + kernel w/ clock/divider options (unless it's new enough to not need it) and to *not* use the VMware Tools host time sync.
That said, you should certainly still have VMware Tools installed, it just sounds like the host time sync is no longer preferred...
Also note that they recommend you remove the local time source in ntp.conf...
Ray
[1] http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd...
I had a similar issue with my Xen VMs, both fully virtualized and para virtualized.
I followed these dirs and was able to fix it.
Perhaps its applicable to you?
http://www.linux.org.za/Lists-Archives/glug-tech-0905/msg00271.html
On Oct 13, 2009, at 4:34 PM, Ray Van Dolson wrote:
On Tue, Oct 13, 2009 at 04:31:03PM -0700, nate wrote:
Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
[..]
VMware Tools not installed.
You should certainly install vmware tools, and enable time sync to the guest. Also don't run an ntp server in a Vmware VM.
This is what I'd always thought, but the VMware KB link[1] referenced in the other reply in this thread seems to indicate that best practice is to use NTP + kernel w/ clock/divider options (unless it's new enough to not need it) and to *not* use the VMware Tools host time sync.
That said, you should certainly still have VMware Tools installed, it just sounds like the host time sync is no longer preferred...
Also note that they recommend you remove the local time source in ntp.conf...
Ray
[1] http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd... _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Tue, Oct 13, 2009 at 4:34 PM, Ray Van Dolson rayvd@bludgeon.org wrote:
On Tue, Oct 13, 2009 at 04:31:03PM -0700, nate wrote:
You should certainly install vmware tools, and enable time sync to the guest. Also don't run an ntp server in a Vmware VM.
This is what I'd always thought, but the VMware KB link[1] referenced in the other reply in this thread seems to indicate that best practice is to use NTP + kernel w/ clock/divider options (unless it's new enough to not need it) and to *not* use the VMware Tools host time sync.
That said, you should certainly still have VMware Tools installed, it just sounds like the host time sync is no longer preferred...
Also note that they recommend you remove the local time source in ntp.conf...
Ray
[1] http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd...
Right. The instructions in the above KB article are contrary to the tactics we used to employ. At least one of my VM guests whose clock was going too fast was corrected by following that KB.
However, there is no one-size-fits-all solution. The article is the best effort by VMWare but it does not rectify all cases. As a matter of fact, the VMWare techs have helped Red Hat improve the RHEL kernel by providing patches (as I mentioned in my earlier post). The details are in this bugzilla:
https://bugzilla.redhat.com/show_bug.cgi?id=463573
Those patches are now in the -164 kernel.
Akemi
On Tue, 13 Oct 2009, Ray Van Dolson wrote:
On Tue, Oct 13, 2009 at 04:31:03PM -0700, nate wrote:
Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
[..]
VMware Tools not installed.
You should certainly install vmware tools, and enable time sync to the guest. Also don't run an ntp server in a Vmware VM.
This is what I'd always thought, but the VMware KB link[1] referenced in the other reply in this thread seems to indicate that best practice is to use NTP + kernel w/ clock/divider options (unless it's new enough to not need it) and to *not* use the VMware Tools host time sync.
That said, you should certainly still have VMware Tools installed, it just sounds like the host time sync is no longer preferred...
Also note that they recommend you remove the local time source in ntp.conf...
Indeed, they changed course over time once they learned that NTP could be made to work reliably when using tinker panic 0. I have had my share of VMware timekeeping troubles the past 5 years, mostly because the recommendations didn't always apply to what we were seeing.
We still use Host-Guest synchronization for ESX 2.0 VM guests, but most of the infrastructure has been migrated, recently to ESX 3.5. VMware never could confirm that the recommendations laid out in the knowledge base article also applies to ESX 2.0.
They seem to update that document (and the timekeeping PDF) for every new ESX release, and removing anything that applied to the previous release :-/ And without a detailed changelog and no access to previous versions of the document you may get paranoid or get into discussions based on different copies of that document.
I've been there too :-)
The CentOS wiki link I mentioned [ http://wiki.centos.org/TipsAndTricks/VMWare_Server ], has instructions adapted from VMware knowledge base link mentioned by many of you [ http://kb.vmware.com/kb/1006427 ].
I have followed the instructions on CentOS wiki, but it doesn't seem to work. Most of the suggestions here are same as mentioned in the CentOS wiki/VMware knowledge base. Any comments or suggestions?
- CS.
On Tue, Oct 13, 2009 at 6:34 PM, Ray Van Dolson rayvd@bludgeon.org wrote:
On Tue, Oct 13, 2009 at 04:31:03PM -0700, nate wrote:
Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
[..]
VMware Tools not installed.
You should certainly install vmware tools, and enable time sync to the guest. Also don't run an ntp server in a Vmware VM.
This is what I'd always thought, but the VMware KB link[1] referenced in the other reply in this thread seems to indicate that best practice is to use NTP + kernel w/ clock/divider options (unless it's new enough to not need it) and to *not* use the VMware Tools host time sync.
That said, you should certainly still have VMware Tools installed, it just sounds like the host time sync is no longer preferred...
Also note that they recommend you remove the local time source in ntp.conf...
Ray
[1] http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd... _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Wed, Oct 14, 2009 at 09:41:52AM -0500, Carlos Santana wrote:
The CentOS wiki link I mentioned [ http://wiki.centos.org/TipsAndTricks/VMWare_Server ], has instructions adapted from VMware knowledge base link mentioned by many of you [ http://kb.vmware.com/kb/1006427 ].
I have followed the instructions on CentOS wiki, but it doesn't seem to work. Most of the suggestions here are same as mentioned in the CentOS wiki/VMware knowledge base. Any comments or suggestions?
(Most people prefer you don't top post FYI).
Can you describe the steps you took exactly? And the symptoms you're still seeing?
Ray
On Wed, Oct 14, 2009 at 9:50 AM, Ray Van Dolson rayvd@bludgeon.org wrote:
On Wed, Oct 14, 2009 at 09:41:52AM -0500, Carlos Santana wrote:
The CentOS wiki link I mentioned [ http://wiki.centos.org/TipsAndTricks/VMWare_Server ], has instructions adapted from VMware knowledge base link mentioned by many of you [ http://kb.vmware.com/kb/1006427 ].
I have followed the instructions on CentOS wiki, but it doesn't seem to work. Most of the suggestions here are same as mentioned in the CentOS wiki/VMware knowledge base. Any comments or suggestions?
(Most people prefer you don't top post FYI).
Can you describe the steps you took exactly? And the symptoms you're still seeing?
Ray _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Following steps were taken to set up my CentOS VM guest (without vmware-tools installed):
1. Kernel options added as: divider=10 clocksource=acpi_pm
2. ntp.conf file was modified as per wiki instructions as follows: (add to the top of the file) # modification as per http://kb.vmware.com/kb/1006427 # The configuration directive tinker panic 0 instructs NTP not to give up # if it sees a large jump in time and must be at the top of the ntp.conf file. tinker panic 0 # end of mod
(comment out 2 lines as below) # modification as per http://kb.vmware.com/kb/1006427 # It is also important not to use the local clock as a time source, # often referred to as the Undisciplined Local Clock. NTP has a # tendency to fall back to this in preference to the remote servers # The following 2 lines commented out.
# when there is a large amount of time drift. # server 127.127.1.0 # local clock # fudge 127.127.1.0 stratum 10
3. Create /etc/ntp/step-tickers and add these lines: 0.centos.pool.ntp.org 1.centos.pool.ntp.org ----------------------
Any missing pointers..?
Thanks, CS.
On Wed, Oct 14, 2009 at 8:55 AM, Carlos Santana neubyr@gmail.com wrote:
Following steps were taken to set up my CentOS VM guest (without vmware-tools installed):
- Kernel options added as:
divider=10 clocksource=acpi_pm
(snip)
Any missing pointers..?
Thanks, CS.
I hope you have seen my earlier posts in this thread. If you are not running the latest kernel (-164), you want to give it a try because it has patches developed by VMWare engineers for RHEL kernels. If you are already running the -164 kernel, you should not be using the clocksource= option as explained in the upstream bugzilla I cited.
Akemi
On Wed, Oct 14, 2009 at 11:04 AM, Akemi Yagi amyagi@gmail.com wrote:
On Wed, Oct 14, 2009 at 8:55 AM, Carlos Santana neubyr@gmail.com wrote:
Following steps were taken to set up my CentOS VM guest (without vmware-tools installed):
- Kernel options added as:
divider=10 clocksource=acpi_pm
(snip)
Any missing pointers..?
Thanks, CS.
I hope you have seen my earlier posts in this thread. If you are not running the latest kernel (-164), you want to give it a try because it has patches developed by VMWare engineers for RHEL kernels. If you are already running the -164 kernel, you should not be using the clocksource= option as explained in the upstream bugzilla I cited.
Akemi
Thanks Akemi.
I am using -128 (2.6.18-128.el5) though.
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Carlos Santana wrote on 10/14/2009 12:13 PM:
On Wed, Oct 14, 2009 at 11:04 AM, Akemi Yagi amyagi@gmail.com wrote:
...
I hope you have seen my earlier posts in this thread. If you are not running the latest kernel (-164), you want to give it a try because it has patches developed by VMWare engineers for RHEL kernels. If you are already running the -164 kernel, you should not be using the clocksource= option as explained in the upstream bugzilla I cited.
Not to get too far OT, but can anybody comment on how this might apply to VirtualBox?
Akemi
Thanks Akemi.
I am using -128 (2.6.18-128.el5) though.
That's the original 5.3 kernel. With 5.4 hitting the streets, and a lot of security/bug-fix patches along the way, one might consider updating.
Phil
On Thu, Oct 15, 2009 at 09:00:54AM -0400, Phil Schaffner wrote:
Carlos Santana wrote on 10/14/2009 12:13 PM:
On Wed, Oct 14, 2009 at 11:04 AM, Akemi Yagi amyagi@gmail.com wrote:
...
I hope you have seen my earlier posts in this thread. If you are not running the latest kernel (-164), you want to give it a try because it has patches developed by VMWare engineers for RHEL kernels. If you are already running the -164 kernel, you should not be using the clocksource= option as explained in the upstream bugzilla I cited.
Not to get too far OT, but can anybody comment on how this might apply to VirtualBox?
Can't comment specifically, except to say that VirtualBox 3.0.8 on my up to date Centos 5.3 box (with a XP guest) seems to allow the guest to keep accurate time.
Well, one of our boxes is running VMware server, and the admin who set it up is under the impression that the NIC *has* to be in promiscuous mode. We, of course, would rather it wasn't.
Is this the case?
mark
m.roth@5-cent.us wrote:
Well, one of our boxes is running VMware server, and the admin who set it up is under the impression that the NIC *has* to be in promiscuous mode. We, of course, would rather it wasn't.
Is this the case?
If your running a bridge at least yes, maybe with other modes too. Bridging is the default mode, and is the most flexible.
nate
m.roth@5-cent.us wrote:
Well, one of our boxes is running VMware server, and the admin who set it up is under the impression that the NIC *has* to be in promiscuous
mode.
We, of course, would rather it wasn't.
Is this the case?
If your running a bridge at least yes, maybe with other modes too. Bridging is the default mode, and is the most flexible.
*sigh*
We were hoping we could lock it down. Thanks.
mark
m.roth@5-cent.us wrote:
We were hoping we could lock it down. Thanks.
Use ESX, the virtual switches in ESX can have promiscuous mode disabled on them, while I suspect the uplinks from the host to the switches have it enabled, the guest VMs cannot access the wire because it's blocked at the virtual switch level.
http://pubs.vmware.com/vsp40_i/server_config/t_edit_the_layer_2_security_pol...
VMware server is really just a toy, I use it myself(at home) but wouldn't use it in any situation where security or performance was important.
nate
On Wed, Oct 14, 2009 at 9:04 AM, Akemi Yagi amyagi@gmail.com wrote:
I hope you have seen my earlier posts in this thread. If you are not running the latest kernel (-164), you want to give it a try because it has patches developed by VMWare engineers for RHEL kernels. If you are already running the -164 kernel, you should not be using the clocksource= option as explained in the upstream bugzilla I cited.
I'm a latecomer to this thread, but I went back and read the whole thing, and my WinXP VMWare Server guest is still way off - more than an hour.
I turned off the VMWare time sync, and also used the control panel to turn off its time sync (wasn't working anyway - errors on every effort), installed meinberg's ntp and ntp sync service, and the time is still off by more than an hour.
I'm not sure what more I need to do to get this working right.
I am running the 164 kernel - have been since about five minutes after I installed it....
Suggestions?
mhr
On Wed, Oct 14, 2009 at 10:55:40AM -0500, Carlos Santana wrote:
Following steps were taken to set up my CentOS VM guest (without vmware-tools installed):
- Kernel options added as:
divider=10 clocksource=acpi_pm
- ntp.conf file was modified as per wiki instructions as follows:
(add to the top of the file) # modification as per http://kb.vmware.com/kb/1006427 # The configuration directive tinker panic 0 instructs NTP not to give up # if it sees a large jump in time and must be at the top of the ntp.conf file. tinker panic 0 # end of mod
(comment out 2 lines as below) # modification as per http://kb.vmware.com/kb/1006427 # It is also important not to use the local clock as a time source, # often referred to as the Undisciplined Local Clock. NTP has a # tendency to fall back to this in preference to the remote servers # The following 2 lines commented out.
# when there is a large amount of time drift. # server 127.127.1.0 # local clock # fudge 127.127.1.0 stratum 10
- Create /etc/ntp/step-tickers and add these lines:
0.centos.pool.ntp.org
1.centos.pool.ntp.org
Any missing pointers..?
Looks good to me. The only issue might be if you have a firewall blocking access to the given NTP servers.
What kind of symptoms are you seeing after doin the above, rebooting, sync'ing the time, then leaving ntpd doing it's thing?
Does the clock start going too fast? Too slow? Do you see any errors in dmesg or /var/log/messages?
Ray
On Wed, Oct 14, 2009 at 10:04 AM, Ray Van Dolson rayvd@bludgeon.org wrote:
On Wed, Oct 14, 2009 at 10:55:40AM -0500, Carlos Santana wrote:
- Kernel options added as:
divider=10 clocksource=acpi_pm
- ntp.conf file was modified as per wiki instructions as follows:
(add to the top of the file) # modification as per http://kb.vmware.com/kb/1006427 # The configuration directive tinker panic 0 instructs NTP not to give up # if it sees a large jump in time and must be at the top of the ntp.conf file. tinker panic 0 # end of mod
(comment out 2 lines as below) # modification as per http://kb.vmware.com/kb/1006427 # It is also important not to use the local clock as a time source, # often referred to as the Undisciplined Local Clock. NTP has a # tendency to fall back to this in preference to the remote servers # The following 2 lines commented out.
# when there is a large amount of time drift. # server 127.127.1.0 # local clock # fudge 127.127.1.0 stratum 10
- Create /etc/ntp/step-tickers and add these lines:
0.centos.pool.ntp.org
1.centos.pool.ntp.org
Any missing pointers..?
Looks good to me. The only issue might be if you have a firewall blocking access to the given NTP servers.
What kind of symptoms are you seeing after doin the above, rebooting, sync'ing the time, then leaving ntpd doing it's thing?
Does the clock start going too fast? Too slow? Do you see any errors in dmesg or /var/log/messages?
Indeed, that is a good question. In /var/log/messages, do you see a line similar to:
<date time> <host> ntpd[6569]: synchronized to xxx.xxx.xxx.xxx
Akemi
Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation: http://wiki.centos.org/TipsAndTricks/VMWare_Server , however it didn't help. I used kickstart for creating this VM and I am listing important steps in ref to timekeeping issue. Any comments or suggestion would be appreciated.
You want to look at VMware's Best Practices for timekeeping. It says to use Linux's NTP, *not* VMware tools.
mark
CS.
# For EL5 virtual machines, Append the following in Grub to help keep the clock from drifting # and to reduce the interupt requests # 32bit: --append="rhgb quiet divider=10 clocksource=acpi_pm" # 64bit: --append="rhgb quiet notsc divider=10" bootloader --location=mbr --md5pass=$1$mXSD1l6mO$BBCk1gYArAATS7dlCQGthN. --append="rhgb quiet divider=10 clocksource=acpi_pm"
%packages --nobase # Other packages not listed here # ntp was installed ntp
### Add step-tickers ### cat > /etc/ntp/step-tickers <<\EOF2 0.centos.pool.ntp.org 1.centos.pool.ntp.org EOF2
### End of step-ticker file ###
# Applied patch mentioned on CentOS page # http://wiki.centos.org/TipsAndTricks/VMWare_Server patch --verbose -b -l -i /root/ntp.patch
VMware Tools not installed.
Thanks, CS. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Tue, 13 Oct 2009, Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation:
The issue I had with time drift was due to running NTP inside the VM. Don't do it. The VM should get it's time through the VMWare tools that are installed on the guest. Once I did this the drift disappeared.
On Wed, Oct 14, 2009 at 1:38 PM, James A. Peltier jpeltier@fas.sfu.ca wrote:
On Tue, 13 Oct 2009, Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation:
The issue I had with time drift was due to running NTP inside the VM. Don't do it. The VM should get it's time through the VMWare tools that are installed on the guest. Once I did this the drift disappeared.
-- James A. Peltier
I recently wrote a script that retrieves the current time from an NTP server and compares it to the local time. You may be surprised to know that inside of a VM the time is constantly drifting, and when it gets to 60 seconds difference, it is forcibly reset by vmware tools. In other words, this is a bad way to handle time.
As has been already mentioned, the timekeeping best practices from vmware have changed from using vmware tools to using NTP and correct kernel options (http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd...). I have not yet implemented this myself, but it seems like a better way to handle it.
Brian Mathis wrote:
On Wed, Oct 14, 2009 at 1:38 PM, James A. Peltier jpeltier@fas.sfu.ca wrote:
On Tue, 13 Oct 2009, Carlos Santana wrote:
Howdy,
I am having time-drift issues on my CentOS VM. I had referred to following documentation:
The issue I had with time drift was due to running NTP inside the VM. Don't do it. The VM should get it's time through the VMWare tools that are installed on the guest. Once I did this the drift disappeared.
-- James A. Peltier
I recently wrote a script that retrieves the current time from an NTP server and compares it to the local time. You may be surprised to know that inside of a VM the time is constantly drifting, and when it gets to 60 seconds difference, it is forcibly reset by vmware tools. In other words, this is a bad way to handle time.
As has been already mentioned, the timekeeping best practices from vmware have changed from using vmware tools to using NTP and correct kernel options (http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd...). I have not yet implemented this myself, but it seems like a better way to handle it.
So is this going to change for CentOS when 5.4 is released?
Also, this seems geared towards guest settings when running under ESX(i). The place I've seen the worst problem is when running under vmware server with a CPU that saves power by running at variable speed.
Les Mikesell wrote:
So is this going to change for CentOS when 5.4 is released?
Also, this seems geared towards guest settings when running under ESX(i). The place I've seen the worst problem is when running under vmware server with a CPU that saves power by running at variable speed
The short answer is "don't do that". ;)
Turn off the 'cpuspeed' init script and any BIOS auto power saving modes. You just aren't going to get accurate time keeping otherwise.
-
Akemi Yagi -
aurfalien＠gmail.com -
Benjamin Franz -
Brian Mathis -
Carlos Santana -
Dag Wieers -
fred smith -
Greg Bailey -
James A. Peltier -
Les Mikesell -
m.roth＠5-cent.us -
mark
-
MHR -
nate -
Phil Schaffner -
Ray Van Dolson