Pootle down

l10n
#1

Hi, Christian and *

Pootle seems down since some hours

http://translations.documentfoundation.org/

The window time for translations is missing and security for job is
being evanescent

#2

Hi *,

Pootle seems down since some hours

Just to let you know - this time virtualbox VM did freeze/slow down to
a crawl, as if it would run on a 1Hz processor or something. I suspect
it has to do with time drift/adjustments and that did screw up the
guest's timers. (host's monitoring shows a rather bigger-than-usual
ntp time-difference - not at the same time when the guest's monitoring
stops, but close enough for me to see a correlation)

Unfortunately neither the Virtualbox Logfiles as the guest's system
logs contain anything unusual (and when it did hang/crawl the box was
still "alive" - ping did work, ssh login worked after loooooong
waiting time, I could even issue a "uptime" command, but I had to type
it blindly, and it took way more than 10 minutes (I don't know, since
I didn't keep watching) to see any reaction, but load on the guest as
well as the host was non-existant, i.e. both were idle, the VM just
was slooooooooooooow)

Since it was still responding, I tried to shut it down cleanly,
sending acpipowerbutton event (and I did see the event in the
ssh-login shell after a while (the "system is going to shut down"
message)) - but After as it was still not shut down at 6am, I had to
kill it (send poweroff). So while I knew about it at the evening, it
was offline till the next morning - sorry for that, but my hope really
was that the shutdown would finish. And when you already waited two
hours, you tell yourself: Just wait a little more, it cannot take
/that/ much longer - at least the vm still responds to a ping, i.e. it
is not dead... And after having waited 4 hours "hey, the network stack
seems down already" (no ping)... But in the end you cannot wait and
wait, and I set myself the limit of having the VM back online in the
morning (CEST)... mysql wasn't happy to have been interrupted (so
mysql was still running.... and I had to use repair table to flag the
tables as clean again)

For the technically inclined: I did set the TSC to be execution-bound
(VBoxInternal/TM/TSCTiedToExecution) - if a freeze reoccurs (symptoms
are that you can still ping, but don't get any repsonse of the regular
webpage, just your browser's timeout message), I'll set a different
clocksource within the guest, as that also shows up in some similar
reports as a workaround for VM guest freezes/hangs.

(If you ask why I set time stamp counter instead of changing the
guest's clock-source: I remember having problems with another VM and
its disk-controllers, where it would reduce the disk i/o speed
dramatically - setting the parameter solved the problem)

So thanks for the downtime-notice (don't hesitate to also ping me on
IRC, to make sure I did notice, but don't forget the mail to this
list, to let others know as well), and sorry for the downtime :-((

ciao
Christian

#3

Many, many thanx for your hard work, Christian.
I know the hard job it is, especially when the system is puzzling one.
If we shall give you a hand... I will do that you say... ping to IRC
and maillist.

At least, I could finish the main part of my job.
:slight_smile: