Have used Centos 5 now couple of weeks and started to find pieces on places, ie. found logs :D
Now, these ntpd errors strances me. Anyone else getting these?
Errors frequency error 500 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 503 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 504 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 505 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 509 PPM exceeds tolerance 500 PPM: 3 time(s) frequency error 512 PPM exceeds tolerance 500 PPM: 203 time(s)
I compiled vanilla kernel 2.6.21.2 into use, just by oldfasion way "make menuconfig, make, make install", used .config from original kernel, came with distro. Only changed prosessor type into K7. Running x86 system, with 512M ram. Some about 200G HD space..
Syncing time into ntp1.funet.fi, but I think that's not the reason, because tried some other servers too, and got somewhat same results...
Despite of errors, all seems to work, or do I imagine that?
Jarmo
On May 24, 2007, at 10:36 PM, jarmo wrote:
Have used Centos 5 now couple of weeks and started to find pieces on places, ie. found logs :D
Now, these ntpd errors strances me. Anyone else getting these?
Errors frequency error 500 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 503 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 504 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 505 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 509 PPM exceeds tolerance 500 PPM: 3 time(s) frequency error 512 PPM exceeds tolerance 500 PPM: 203 time(s)
That means that your clock is drifting too badly for ntpd to compensate by slewing time, and it's had to step instead. So time on your machine is discontinuous and possibly stepping backwards. You can find some more information by searching for "500" in the ntpd manpage.
I think this would be due to bad hardware or a kernel problem.
I compiled vanilla kernel 2.6.21.2 into use, just by oldfasion way "make menuconfig, make, make install", used .config from original kernel, came with distro. Only changed prosessor type into K7.
You might have done this to yourself by switching kernels. (I've heard of time problems due to a problem with a SATA controller driver.) What was wrong with the one that came with CentOS? If you switch back to it, do you still have this problem?
Syncing time into ntp1.funet.fi, but I think that's not the reason, because tried some other servers too, and got somewhat same results...
Despite of errors, all seems to work, or do I imagine that?
It won't violently explode, but time going backwards can cause problems. For example, "make" can produce incorrect rebuilds due to improper ordering of timestamps.
--On Thursday, May 24, 2007 10:54 PM -0700 Scott Lamb slamb@slamb.org wrote:
It won't violently explode, but time going backwards can cause problems. For example, "make" can produce incorrect rebuilds due to improper ordering of timestamps.
Will security-sensitive apps like sshd complain and refuse access, assuming some kind of exploit is in progress?
On May 25, 2007, at 8:17 AM, Kenneth Porter wrote:
It won't violently explode, but time going backwards can cause problems. For example, "make" can produce incorrect rebuilds due to improper ordering of timestamps.
Will security-sensitive apps like sshd complain and refuse access, assuming some kind of exploit is in progress?
Deliberately breaking? Maybe, but not that I know of. There are some things that won't work if the clock's just too far off - like Kerberos issues tickets valid for five minutes, so a delta between the KDC and the host of more than that will cause problems.
On Thu, 24 May 2007, Scott Lamb wrote:
On May 24, 2007, at 10:36 PM, jarmo wrote:
Have used Centos 5 now couple of weeks and started to find pieces on places, ie. found logs :D
Now, these ntpd errors strances me. Anyone else getting these?
Errors frequency error 500 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 503 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 504 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 505 PPM exceeds tolerance 500 PPM: 1 time(s) frequency error 509 PPM exceeds tolerance 500 PPM: 3 time(s) frequency error 512 PPM exceeds tolerance 500 PPM: 203 time(s)
That means that your clock is drifting too badly for ntpd to compensate by slewing time, and it's had to step instead. So time on your machine is discontinuous and possibly stepping backwards. You can find some more information by searching for "500" in the ntpd manpage.
I think this would be due to bad hardware or a kernel problem.
Or something like a saturated I/O subsystem.
I've got an x86_64 CentOS 5 server with two SATA drives, mirrored with software RAID 1. The auditctl ruleset was causing auditd to log like crazy, and ntpd was paying the price.
With all the logging going on, the drift was bumping up against the 500 mark. Once I altered the ruleset, the drift has settled down to a reasonable single-digit number.
So you might check to see if top or iostat is reporting a lot of disk usage...
Scott Lamb kirjoitti viestissään (lähetysaika perjantai, 25. toukokuuta 2007 08:54):
I think this would be due to bad hardware or a kernel problem.
You might have done this to yourself by switching kernels. (I've heard of time problems due to a problem with a SATA controller driver.) What was wrong with the one that came with CentOS? If you switch back to it, do you still have this problem?
Tested and didn't get those errors. Then compiled 2.6.21.3 vanilla, as thought, it was kernel problem, got those errors again. Seems that Athlon has some kind of troubles with new kernels. Namely in original centos kernel there is pentium pro optimized...
Ok, next step was, I stopped cpuspeed from services, edited ntp's drift file into 0.000 and restarrted ntpd, no errors after that. Time is well in sync and not mentionable drifting in sight...
May 27 06:40:13 oh1mrr ntpd[4986]: synchronized to LOCAL(0), stratum 10 May 27 06:40:23 oh1mrr ntpd[4986]: synchronized to 193.166.5.177, stratum 2
Jarmo
-
jarmo -
Kenneth Porter -
Paul Heinlein -
Scott Lamb