I use mirror.pl to create a local copy of the openSUSE repos. I then use RSYNC to copy to a secondary server. I have noticed that the time stamps on the secondary server do not match my main download copy.
I wrote a script that compares timestamps between my servers.
i586/dhcp-devel-4.2.4.P2-0.1.4.1.i586.rpm
Left Right
Birth 0 0
Modified 1351012521 1348562700 Differ
Accessed 1348562700 1348562700
Changed 1351012521 1350040218 Differ
Modified Human 2012-10-23 13:15:21.000000000 -0400 2012-09-25 04:45:00.000000000 -0400
MD5Sum 8b382a45dc5181df741f103401e917f6 8b382a45dc5181df741f103401e917f6
I even tried to add --times to the end of the command. I tested by deleting the 12.2 OSS Updates directory and running the rsync again.
I did not see anything in the man page that says that -t/–times only works when the RSYNC demon is running. I know that “Changed” is the entry in the file system and RSYNC will not preserve that. But RSYNC is preserving Accessed time, so how come it is not preserving Modified time?
I have never had problems with time and rsync, including when I was using 11.4.
That said, the time sync does not work if the destination file system is NTFS, unless the process doing the update has sufficient permissions. That’s an NTFS restriction, not an “rsync” problem.
I’m not sure about other file systems. As far as I know, you should be fine with standard unix file systems.
On my own computers, when copying to NTFS, I must either do it as root, or I must use mount options so that I am the file owner, if I want to be able to set the time.
Check what type of file system is being used on the destination.
My questions would be: 1) Why use XFS over EXT4? What problem does the time difference cause? Have you considered trying different PC’s to see if you get the same result? Ether the problem is the process you follow or the file system that you have selected or there is no problem because it would do the same thing everywhere.
On 2012-10-24 23:16, jdmcdaniel3 wrote:
>
> dwestf;2498551 Wrote:
>> The file system is XFS on both sides.
>
> My questions would be: 1) Why use XFS over EXT4? What problem does the
> time difference cause? Have you considered trying different PC’s to see
> if you get the same result? Ether the problem is the process you follow
> or the file system that you have selected or there is no problem because
> it would do the same thing everywhere.
The filesystem will not be the cause, as long as it is a Linux
filesystem. If it were, it would be a severe bug.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” (Minas Tirith))
1, Two reasons, one I have always found that an XFS partition repairs itself faster then ext3, two don’t have to wait hours for the system to check itself if I reboot it in two months. Have not tested ext4 yet.
This problem started when I was seeing a large number of files being downloaded from mirror.anl.gov using mirror.pl, but was only seeing a few updates. I don’t have RSYNC access, firewall rules.
I had problems with the openSUSE mirror before, different versions of rpm were being introduced into the mirror and not matching repodata. I wrote a oneliner that would compare the RPM’s on the secondary copy server to the main download. This would tell me what RPM’s were replaced in the repo.
So using this, I was not see any difference in the RPM’s that were being downloaded. That is when I noticed that timestamps were changing at anl.gov. Drafting an email to anl.gov right now. I am just trying to get all copies to match up.
I wounder if this is the same problem as before but instead of different versions it is the same file but different time stamps. I just did a mirror -T and one of the files that got a timestamp update was flashplayer from April. Why would an RPM from April change its time stamp today. I ran mirror -T twice yesterday.
Found an error in my timestamp script. I transposed X and Y in the stat command and was switching access and modified time. So what I am seeing between my two servers is that access time is different. I guess rsync does not preserve access time.
dwestf wrote:
> Found an error in my timestamp script. I transposed X and Y in the stat
> command and was switching access and modified time. So what I am seeing
> between my two servers is that access time is different. I guess rsync
> does not preserve access time.
FWIW, I mount all my data filesystems with noatime anyway Can speed
things up quite a bit, depending on what you’re doing.