Communication of a20140114.1

Datum: Tue, 21 Jan 2014 07:13:15 +0100

Dear Eva,

Thanks for handling this arbitration case.

> As far as I know you issued the mail because there was no free disk
> space left on the system the mail server is running on, so the mail
> server could not work anymore. Is this correct?

This is correct. I got a phone call from [representant of C], who in turn got user
complaints about error messages coming up while trying to send mails.
Because I am not aware of any monitoring of this machine (at least none
that sends me automated error messages to a mobile device or something
that can keep me informed when I'm not using the internet) and there's
no shift planning for email admins, none of us could forsee when or
whether this will happen.

> Can you give an estimate about the importance to free more disk space in
> the near future? (How much free space do we have? What do you think how
> long it will take to fill up that space again?)

The current disk usage is as follows:

Filesystem            Size  Used Avail Use% Mounted on
tmpfs                 2.0G     0  2.0G   0% /lib/init/rw
tmpfs                 2.0G     0  2.0G   0% /dev/shm
rootfs                 55G   53G  1.2G  98% /

The last disk space incident we had was some months ago (for sufficently
precise values of "some", I have to dig in my emails for a while), so I
tend to say, we're not in immediate danger. In order to get a better
feeling  how far we're running out of space, I set up a small cron job
that logs the current disk usage every night at 12 am. Please note that
this is not a replacement for a real monitoring system.

> Do you have any idea when the mail server will be moved to the new
> server? (And the problem will be solved for some time.)

Please ask [other email admin] about this. As [representan of C] will confirm, [other email admin] and I agreed
upon the following strategy a couple of months ago:

1) [other email admin] has a very detailed idea of how the future system should look
like, so he will set up a POC system and will document every step.

2) After the POC system turned out to work properly, I will take [other email admin]'s
document and will try a second setup. If the second system works the
same way as the POC server, we can be sure to have a setup that we know
precisely and can reproduce whenever needed. Every future change on this
system should be going through a proper change management so we can be
sure that there is no hidden cofiguration that no one is really aware of.

> Are there any other - easy - ways to free a relevant amount of disk
> space for the mail server?

Regrettably I already checked when I tried to free space and found
nothing really promising besides deleting mail. Please find below the
current disk space usage on /:

[other directories]

16G    /home

[other directories]

As you can see, most space is being consumed in /home. I already checked
whether we can delete some old logs in /var, but there is no stuff older
than one week, and besides that it does not consume that much space

> Can you give an estimate how much disk-space was freed by the deletion
> of support mails at 2014-01-14?

A really rough guess, because I didn't take notes: Less than 1 GB. Here
are the current home directory sizes sorted by KB. Please handle this
information with care, because it's personal data, although not really
sensitive one.

list of all other home directories - all of them use less space than support
5762568    /home/support

> Do you know if there is a direct backup to the inbox of support other
> than the OTRS?

None that I'm aware of.

> Do you see any reasons to leave - or delete - the remaining mails in the
> mailbox - beside of disk space?

For archiving or legal purposes it might be important to store relevant
mails somewhere. Regrettably a production system's hard drive is only a
makeshift (I guess that's what your last question is heading at). As far
as I know, if we loose this data, we can only rely on the personal
copies the arbitrators made. So it would be a good idea to have some
kind of archive somewhere, but I'm afraid the mail server is not the
best place. Above that, a monitoring system might turn out helpful for
incident handling and capacity planning (although, as I'm fully aware,
it might also be a bit over the top ;)

Hopefully this answers at least some of your questions. Please feel free
to ask me for further details.

CARS and best regards,

Arbitrations/a20140114.1/Communication (last edited 2016-09-01 21:15:36 by EvaStöwe)