#hoopla

One of my ECL programs suddenly stopped working. Whenever I started the workunit, the status stayed "blocked" and it timed out eventually.

hpcc-run.sh status

showed that mythor was stopped. Restarting it didn't help. The start of mythor also timed out. I found helpful error messages in /var/log/HPCCSystems/cluster/cc_hpcc-init_status_*.log. In my case I found

...
xyz.xyz.xyz.41: Running sudo /etc/init.d/hpcc-init start 
tee: /tmp/hpcc_status_20160629_104057_105636: Read-only file system
...

The issue was a hardware defect causing one node in the cluster to switch to a read-only filesystem. After fixing the defect, I could start mythor again.

Set the configuration of the Job to:

conf.set("mapreduce.input.fileinputformat.split.minsize", "" + Long.MAX_VALUE);

Hadoop required individual Key or Value classes to come with a constructor without parameters, such as GroupJoinVal() {}.

Das Problem scheint mit dem Update KB3097877 zusammenzuhängen, welches man unter Systemsteuerung > Installierte Updates anzeigen > Rechter Mausklick darauf > Deinstallieren wieder entfernen kann.

Damit es nicht beim nächsten Updatelauf wieder installiert wird kann man es ausschließen in der Systemsteuerung > Nach Updates suchen > rechter Mausklick auf dem Update > ausblenden.

I changed the mail.db.idle_limit from 300000 to 30000000 as described above. This solved the problem for me on Ubuntu 14.04. It seems to be connected to very large mail folders.