Bugzilla – Bug 113610
Time in logfile messages in UTC for linux kernel entry
Last modified: 2005-08-30 13:56:41 UTC
Each Linux kernel entry has a UTC time, other entries are in localtime.
Created attachment 47882 [details] /var/log/messages
All timestamps in the system log are generated by syslogd, not the kernel itself. So it's a syslogd problem I assume.
Yes. Timestamps for kernel messages are generated by klogd (not by syslog*): 7020 write(1, "<6>Aug 29 13:33:52 kernel: subfs"..., 37) = 37 7020 time([1125315232]) = 1125315232 7020 write(1, "<6>Aug 29 13:33:52 kernel: klogd"..., 76) = 76 7020 read(0, "<7>ISO 9660 Extensions: Microsof"..., 262143) = 49 7020 time([1125315232]) = 1125315232 7020 write(1, "<7>Aug 29 13:33:52 kernel: ISO 9"..., 73) = 73 7020 read(0, "<7>ISO 9660 Extensions: RRIP_199"..., 262143) = 35 7020 time([1125315232]) = 1125315232 7020 write(1, "<7>Aug 29 13:33:52 kernel: ISO 9"..., 59) = 59 7020 read(0, "<6>xen-br0: port 2(peth0) enteri"..., 262143) = 50 7020 time([1125315250]) = 1125315250 7020 write(1, "<6>Aug 29 13:34:10 kernel: xen-b"..., 74) = 74 7020 read(0, "<6>device peth0 left promiscuous"..., 262143) = 88 7020 time([1125315250]) = 1125315250 7020 write(1, "<6>Aug 29 13:34:10 kernel: devic"..., 62) = 62 7020 time([1125315250]) = 1125315250 7020 write(1, "<6>Aug 29 13:34:10 kernel: xen-b"..., 74) = 74 7020 read(0, "<6>ACPI: PCI interrupt for devic"..., 262143) = 56 7020 time([1125315250]) = 1125315250 7020 write(1, "<6>Aug 29 13:34:10 kernel: ACPI:"..., 80) = 80 There is a difference of 12 hours. Is your local time in "M" or "Y" timezone (+/- 12 hours) ? Aug 28 05:58:23 linux kernel: ISO 9660 Extensions: RRIP_1991A Aug 28 05:58:52 linux kernel: ISO 9660 Extensions: Microsoft Joliet Level 3 Aug 28 05:58:52 linux kernel: ISO 9660 Extensions: RRIP_1991A Aug 28 17:59:47 linux shadow[5945]: group already exists - group=ntadmin, by=0 Aug 28 18:00:43 linux shadow[5980]: group is unknown - group=wwwadmin, by=0 Aug 28 18:00:43 linux shadow[5981]: default group changed - account=wwwrun, uid= Aug 28 18:00:43 linux shadow[5982]: shell changed - account=wwwrun, uid=30, shel Aug 28 06:03:02 linux kernel: ISO 9660 Extensions: Microsoft Joliet Level 3 Aug 28 06:03:02 linux kernel: ISO 9660 Extensions: RRIP_1991A Aug 28 06:03:26 linux kernel: ISO 9660 Extensions: Microsoft Joliet Level 3 Aug 28 06:03:26 linux kernel: ISO 9660 Extensions: RRIP_1991A I've installed Beta3 today and there is still some bug in YaST2 clock setting pages. The clock in YaST displayed "localtime +2" with TZ setting as localtime (it seems YaST assumed, that the hwclock is set to UTC). The clock on this PC is set to localtime, not UTC, because of WinXP installed there. Switching the TZ combo box to UTC and back displayed the correct local time. It SEEMS, yast2 restarts only syslogd/-ng (manually, instead via init script) and forgets to restart klogd as well, after it has changed the TZ settings. Or something like this... As result, the klogd runs with old TZ settings and syslogd/-ng with new... I would say, there is nothing we can fix in syslog packages. YaST seems to need a fix.
This has nothing todo with syslogd nor with klogd.
My HW-Clock is set to the localtime. Region: Pacific Time Zone: New Zealand Hardware Clock set to: Local Time After several cold starts the time of all entries in logfile are always localtime. It seems to be only a problem in the installing procedure.
The clock detecting/setting problem mentioned by Marius should be fixed in beta4. However, after setting the time, YaST does not restart neither syslogd nor klogd. It only sets the time using hwclock. I don't know if hwclock itself restarts logging services - does anyonw know?
No, hwclock does only sets the date. I've tested a little bit with syslog-ng. There is no need to restart it neither on timezone nor on time/date changes, since the glibc functions reads the /etc/localtime every time: root@Xanthos:~ # logger "======= " $(date) " ============" ; date Di Aug 30 14:29:52 CEST 2005 root@Xanthos:~ # date 08301230 #MMDDhhmm Di Aug 30 12:30:00 CEST 2005 root@Xanthos:~ # logger "======= " $(date) " ============" ; date Di Aug 30 12:30:06 CEST 2005 root@Xanthos:~ # cp -a /usr/share/zoneinfo/UTC /etc/localtime root@Xanthos:~ # date Di Aug 30 10:32:12 UTC 2005 root@Xanthos:~ # logger "======= " $(date) " ============" root@Xanthos:~ # cp -a /usr/share/zoneinfo/Europe/Berlin /etc/localtime root@Xanthos:~ # logger "======= " $(date) " ============" root@Xanthos:~ # date Di Aug 30 12:33:05 CEST 2005 root@Xanthos:~ # rcntp restart Shutting down network time protocol daemon (NTPD) done Try to get initial date and time via NTP from 192.168.110.200 192.16done0.254 Starting network time protocol daemon (NTPD) done root@Xanthos:~ # date Di Aug 30 14:34:34 CEST 2005 root@Xanthos:~ # logger "======= " $(date) " ============" Aug 30 14:29:52 Xanthos root: ======= Di Aug 30 14:29:52 CEST 2005 =========== Aug 30 14:29:57 Xanthos sshd[5710]: Accepted publickey for root from 127.0.0.2 p Aug 30 12:30:06 Xanthos root: ======= Di Aug 30 12:30:06 CEST 2005 =========== Aug 30 10:32:18 Xanthos root: ======= Di Aug 30 10:32:18 UTC 2005 ============ Aug 30 12:33:02 Xanthos root: ======= Di Aug 30 12:33:02 CEST 2005 =========== Aug 30 14:34:23 Xanthos ntpdate[5851]: step time server 192.168.110.200 offset 7 Aug 30 14:34:23 Xanthos ntpd[5856]: ntpd 4.2.0a@1.1191-r Fri Aug 19 22:07:11 UTC Aug 30 14:34:23 Xanthos ntpd[5856]: precision = 1.000 usec Aug 30 14:34:23 Xanthos ntpd[5856]: Listening on interface wildcard, 0.0.0.0#123 Aug 30 14:34:23 Xanthos ntpd[5856]: Listening on interface wildcard, ::#123 Aug 30 14:34:23 Xanthos ntpd[5856]: Listening on interface lo, 127.0.0.1#123 Aug 30 14:34:23 Xanthos ntpd[5856]: Listening on interface eth0, 192.168.110.1#1 Aug 30 14:34:23 Xanthos ntpd[5856]: kernel time sync status 0040 Aug 30 14:34:38 Xanthos root: ======= Di Aug 30 14:34:38 CEST 2005 =========== As far as I can see, klogd uses a time(&now), ctime(&now) combination what should be also OK. Hmm.. is there some chroot used and the log forwarded between chroot and the system? A chroot would explain it: this will happen if the /etc/localtime of the system and $chroot/etc/localtime differs and one of the daemons runs in a chroot...
(In reply to comment #5) > It seems to be only a problem in the installing procedure. Yes, in normal system this does not happen.
So if I understand it correct, it should be fixed by the fix of the bug already mentioned in comment #3 (... "it seems YaST assumed, that the hwclock is set to UTC"...).