On Fri, May 30, 2008 at 09:27:51AM +0300, Volkan YAZICI wrote: > This morning one of our R&D servers stop responding (no ssh, http) and > because of urgency of some tests I needed to hardware-reset it. After > machine woke up, I first checked /var/log/messages: > [snip most] > May 30 08:09:47 arge -- MARK -- > May 30 08:29:47 arge -- MARK -- > May 30 08:44:36 arge kernel: e100: eth1: e100_watchdog: link down > May 30 08:44:38 arge kernel: e100: eth1: e100_watchdog: link up, 100Mbps, > full-duplex > May 30 08:44:42 arge kernel: e100: eth1: e100_watchdog: link up, 100Mbps, > full-duplex
> May 30 08:45:14 arge shutdown[7450]: shutting down for system halt > May 30 08:38:11 arge syslogd 1.4.1#18: restart. > > As can be understood from "kernel: e100: eth1: ..." lines, I first > suspected a connection failure and try to fiddle with the network cable > socket. But logs tell that it wasn't the problem. Moreover, it seems > that system was working properly just before 08:44:36 if we'd look at > /var/log/syslog > [snip] > I checked logs of every file under /var/log at time between 08:00:00 and > 08:38:00, but found nothing useful. OTOH, if we'd look at below lines of > the /var/log/messages output: > > May 30 08:45:14 arge shutdown[7450]: shutting down for system halt > May 30 08:38:11 arge syslogd 1.4.1#18: restart. > > It seems that openntpd somehow failed to synchronize hardware clock with > the time it gathered from NTP servers, and after reboot it switched back > to a past time. Is this something expected? If not, how can I fix this? > > To summarize, what else should I check to figure out the reason of the > emerged problem? (I'll try to login from terminal next time such a > failure repeats.) I don't know what caused the freeze; The hard reset would keep the shutdown scripts from setting the system time to the hardware clock. On restart, did the ntpd eventually get a network connection and fix the time? It may not have been a freeze at all, just a networking problem that wasn't found by fitzing with the cable. Logging in from a VT or serial terminal would have been helpful. If you are concerned that this may happen again, you may even want to connect up a serial console to another box (or a real serial VT) and watch that as well. Doug. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]