[server-admins] [Server-admins] CrossWire Server -- Admin Status Summary 20090819

[Jonathan Marsden]

Back on 25 July 2009, Troy A. Griffitts wrote:

> CrossWire's server configuration currently sucks.

This is my quick look at where things now stand on the issues raised at
that time, plus two others that emerged during discussion of them:

(1) UPDATING MEDIAWIKI

> o    when we get mediawiki updates from our occasional yum update, the
> wiki stops working (like right now).

This has been identified as being caused by the need to run an update
script /usr/share/mediawiki/maintenance/update.php after such upgrades
to the mediawiki software.

However, this is not known to be 100% safe, and a backup before doing
this is recommended.  Therefore, a mediawiki backup script and an
associated wrapper that backs up mediawiki, runs yum update, and then
runs the update.php script has been written.

The backup script has been fairly well tested (we have not yet done a
recovery drill to see if the backups it makes are indeed sufficient and
capable of being restored from!).  The update wrapper script is (as far
as I know) still untested.

Currently I (Jonathan) make unofficial mediawiki backups nightly, and
rsync them to my local desktop PC "most days".

NEXT STEPS:

(1A) Testing of the yum-update-and-mediawiki-update.sh script.

(1B) Integration of these scripts into normal CrossWire server admin
processes and procedures.

(2) EMAIL / MAILING LIST ISSUES

> o    mailman does not send admins admin notices
> 
> o    mailman constantly unsubs me because of excessive bounces
> 
> o    to get this new list to actually work, I had to find and fix this
> error in /var/log/exim/main.log ...

These have been found to be caused by two separate interacting
configuration issues in exim.conf , and both are now believed fixed.

> o    our smtp server seems to get bogged down with spam or something and
> rejects valid mail.  Got this response yesterday and have had people ask
> me in the past if my mail account at crosswire was still valid because
> of similar events.
> 
> <scribe at crosswire.org>: 451 Timeout trying verify RCPT
> (scribe at crosswire.org) for domain (crosswire.org)

This was probably (?) also related to the exim config issues and is also
now believed fixed.

NEXT STEPS:

(2A) Monitor exim and mailman logs for issues like these.

(2B) Consider customizing logwatch configs, to make its reports more useful.

(3) HIGH TOMCAT CPU and RAM USAGE

This has been deemed low priority, but during its discussion the
location of the patches which seem to trigger it were rediscovered:

> Looks like http://bugs.sun.com/view_bug.do?bug_id=6567407 is the one.
> That points right back to
> 
>   Source: http://crosswire.org/~scribe/corba-utf8-src.zip 

>   Source patch: http://crosswire.org/~scribe/CORBA-lenient-utf-8-default.patch

NEXT STEPS: No immediate action needed, low priority since we have so
much RAM and so many cores available :)

(4) FORUM SOFTWARE UPDATE

Mannfred has updated the forum software, bringing new features to them
in the process.  This work seems to have been completed successfully.

NEXT STEPS: None needed.

(5) BACKUPS

Troy asked for comment on the current backup strategy, and got a lengthy
response from me :)

NEXT STEPS: Awaiting Troy's response to my comments on this topic.

QUESTION:  What items are missing from this list?  What CrossWire server
admin issues remain unaddressed?

Jonathan