Closed GoogleCodeExporter closed 8 years ago
Ok, basics to check:
Is disk full?
Are services running (mysql, syslog-ng, searchd)?
What is in /data/elsa/log/node.log?
Original comment by mchol...@gmail.com
on 9 Jan 2013 at 7:39
The data dir's disk is an iSCSI mount to a 1.5 TB volume. df -h says it has
1.4T free.
MySql.service is running on the node instance, as are syslog-ng and searchd.
(At least, systemd claims they are. Their respective ports are also open and
accepting data).
My datadir is /srv/, but the node.log file seems to be 18 GB big... Would it be
better to just give you the last 10,000 lines or so? I've attached a file with
the last 10,000 lines.
Original comment by i...@pingas.org
on 12 Jan 2013 at 7:23
Attachments:
Ok, it looks like some process is going haywire and continually reindexing.
Shut down syslog-ng and make sure there are no Perl processes running. It's
also odd that Sphinx is using /etc/sphinx.conf. Did you move the location from
/usr/local/etc/sphinx.conf, which is the ELSA default?
After syslog-ng is shut down and no Perl processes are running, run:
cd /usr/local/elsa/node && echo "" | perl elsa.pl -o
That should complete one full ELSA batch cycle and return with no errors or
warnings.
Original comment by mchol...@gmail.com
on 12 Jan 2013 at 9:36
I'm pretty sure I copied over the /usr/local/etc/sphinx.conf file to
/etc/sphinx.conf (and modified configs accordingly), since I wanted to have
sphinx's conf in the upstream location. I've attached it. If it looks totally
wrong, I'm fine with changing it to whatever it needs to be.
Syslog-ng.service is now shutdown. By the way, my ELSA dir is /usr/share/elsa.
Sorry to be so non-standard here!
I've run the command (with the cd path changed), and the output is below:
--------------------------------------------------------------------------
root@loki ~ # cd /usr/share/elsa/node && echo "" | perl elsa.pl -0
Cannot read config file: /etc/elsa_node.conf at
/usr/share/perl5/site_perl/Config/JSON.pm line 49, <DATA> line 429.
Config::JSON::__ANON__('Config::JSON=HASH(0xa4c78f0)', '/etc/elsa_node.conf') called at constructor Config::JSON::new (defined at /usr/share/perl5/site_perl/Config/JSON.pm line 668) line 41
Config::JSON::new('Config::JSON', '/etc/elsa_node.conf') called at elsa.pl line 52
--------------------------------------------------------------------------
I realized that my elsa configs are in /etc/elsa/ and not just /etc/, so I
linked /etc/elsa/elsa_node.conf to /etc/elsa_node.conf and re-ran that command.
It took much longer to run and the output is below:
--------------------------------------------------------------------------
root@loki elsa/node # cd /usr/share/elsa/node && echo "" | perl elsa.pl -o
Validating directory...
Use of uninitialized value $db_size in concatenation (.) or string at
/usr/share/elsa/node/Indexer.pm line 156.
Use of uninitialized value $db_size in addition (+) at
/usr/share/elsa/node/Indexer.pm line 175.
Use of uninitialized value $db_size in concatenation (.) or string at
/usr/share/elsa/node/Indexer.pm line 191.
Use of uninitialized value $db_size in addition (+) at
/usr/share/elsa/node/Indexer.pm line 211.
DBD::mysql::st execute failed: Table 'syslog.host_stats' doesn't exist at
/usr/share/elsa/node/Indexer.pm line 253.
DBD::mysql::st execute failed: Table 'syslog.host_stats' doesn't exist at
/usr/share/elsa/node/Indexer.pm line 253.
--------------------------------------------------------------------------
I just logged into the web frontend, but the tooltip still says the latest
Index and Archive entries are from 2012-12-31. What's the next step? Should I
just re-install elsa completely?
Original comment by i...@pingas.org
on 14 Jan 2013 at 4:06
Attachments:
Just realized -- that "-0" was actually a "-o" when I ran it. Not sure how it
was changed in my comment. Sorry about that.
Original comment by i...@pingas.org
on 14 Jan 2013 at 4:58
Ok, good, we see the root of the problem "DBD::mysql..." That error indicates
that you haven't run the update, so you need to run "sh
/usr/share/elsa/install.sh node update" to get that table installed so the rest
of the code works. No need to re-install.
Original comment by mchol...@gmail.com
on 15 Jan 2013 at 1:16
Closing due to inactivity. Please comment to reopen.
Original comment by mchol...@gmail.com
on 8 Apr 2013 at 2:05
Commenting just to say that it was eventually solved with an upgrade and a
re-roll of the indexes. However, I'm having different problems now. I'm
reopening a ticket to explain them.
Original comment by i...@pingas.org
on 13 May 2013 at 5:48
Original issue reported on code.google.com by
i...@pingas.org
on 9 Jan 2013 at 7:26