Re: failed to get process data

[Luca Cazzaniga]

the kernel version is 3.10.0-1160.42.2.el7.x86_64 and the script to
stop the problematic service takes care to delete the pidfile
normally.. I couldn't check what happens in that occasion anyway

On Mon, 22 Nov 2021 at 13:39, Luca Cazzaniga <fusillator@gmail.com> wrote:
>
> sorry i miss to say I also double check the load of the server on
> grafana, and it was not on heavy load for what I can see..
> We will evaluate to upgrade monit release if it would be helpful to
> solve the issue
>
> On Mon, 22 Nov 2021 at 13:38, Luca Cazzaniga <fusillator@gmail.com> wrote:
> >
> > Hi Lutz, the pid was saved in a bin directory on a xfs fs, the pidfile
> > was removed by the script of the application server, or at least I
> > suppose so, in that occasion there's been a service stop via monit and
> > the daemon log reported the service was successfully terminated.
> > After that there's been an upgrade of the os, a restart of the server
> > and at the boot monit should have started the daemon again but it
> > looped with the error "sistd failed to get process data" for one
> > application server. Anyway it did its work for other services
> > configured with pidfile without problem.
> > Really strange..
> > Not sure how procfs is updated/flushed, it's an interface to the
> > kernel structures for process statistics written on a cache.. i wonder
> > if it could be deleted or unavailable for some reason..
> >
> > On Mon, 22 Nov 2021 at 08:21, Lutz Mader <lutz.mader@freenet.de> wrote:
> > >
> > > Hello Luca Cazzaniga,
> > > thanks for your answer, response. I can not reproduce my problems too,
> > > but the problem will came back from time to time, on some of my systems.
> > >
> > > > In our case the server was just restarted and the problem had persisted 
> > > > all
> > > > over the week end. The annoying part is that no alert was triggered 
> > > > also if
> > > > the process wasn't actually running... I don't think the server was
> > > > overloaded upon restart anyway but I can't reproduce the problem 
> > > > neither...
> > > > Maybe an event should be launched in that kinds of situation. Just to 
> > > > let
> > > > the sysop know
> > >
> > > Where are the pid files stored, is a tmpfs used?
> > > Do you remove old pid files?
> > > Your Linux version is?
> > >
> > > This is an old version of Monit, any plan to use a new one?
> > >
> > > >> The monit release on the server is 5.25.3
> > >
> > > With regards,
> > > Lutz
> > >