[Users] "Start the VPS" frequently required from web admin page since CentOS 7.2 migration

Scott Dowdle dowdle at montanalinux.org
Fri Mar 4 11:43:53 PST 2016


Greetings,

----- Original Message -----
> Hi Everyone,
> 
> We use a hosting provider to host a website and wiki. Last week we
> performed the CentOS 7.1 -> 7.2 migration. We issued 'sudo yum
> update'
> and everything migrated cleanly. The process seemed to go well.
> 
> Since the upgrade it appears the VPS service frequently stops or
> crashes. We're not sure which at the moment because we don't have
> access to Dmesg and Syslog (more correctly, they are 0-bytes). We
> just
> noticed the trend, but it seems to be about a daily occurrence.
> 
> I've searched for the symptoms, but I'm not getting hits in the
> context of an issue. Most results are related to creating a container
> and starting the server.
> 
> My questions are:
> 
>  * Is there a well known issue that someone can point me to?
>  * Is there an alternate set of log files available for
>  troubleshooting?
>  * How do I troubleshoot in the absence of logs?
> 
> My apologies for the lack of information. I know its not very
> helpful.
> 
> Thanks in advance.

There is no well known issue that I'm aware of.

The container will probably not see kernel related messages because the kernel is on the host node and logging there.

There are only a handful of logs so look around.  I'm guessing there isn't going to be anything of note to be found inside of the container.  I assume your container is using a vSwap config rather than a UBC style.  You can check your /proc/user_beancounters within the container to see if there are any failcounts.  If there are, that'd be something to look into.

You might also consider using journald provided by systemd.  It may more may not be on by default and EL7 has it in non-persistant mode by default I think... and I'm not even sure what the defaults might be in whatever OS Template was used to create your container with.  Maybe your hosting provider uses the stock OS Templates provided by the OpenVZ Project, maybe not.  In any event, journald with peristant storage rocks because you have a single binary log that that can be interacted with like a database once you learn how to use journalctl.  The whole systemd/journald/journalctl thing is really a tanget with regards to the problem you are asking about but maybe that will interest you for future logging info.

The people to contact would be, and I'm sure I'm not telling you anything you don't already know, the hosting provider... to have them troubleshoot for you... since they had root access to the host node.  They may have had some issues rather than it being specific to your particular container.

TYL,
-- 
Scott Dowdle
704 Church Street
Belgrade, MT 59714
(406)388-0827 [home]
(406)994-3931 [work]


More information about the Users mailing list