[GE users] Database Corrupted very often

Adrián Bravo Navarro adrian.bravo at fdi.ucm.es
Mon Oct 27 13:59:10 GMT 2008

Hi guys,

we've been using SGE for a long while now, and sometimes the db gets
corrupted (we dunno exactly why). When this occurs we use to reinstall
sge and it works for a while. At this time, it is happening really
frequently, and we wonder if there is anything we can do, or any config
param we may change to avoid this errors.

These are the log entries related to the errors:

10/23/2008 17:20:59|qmaster|xxxx|E|couldn't close database
environment: /aplicaciones/sge6.1u4//default/spool/spooldb: (22) Invalid
10/23/2008 17:20:59|qmaster|xxxx|E|shutdown of rule "default rule" in
context "berkeleydb spooling" failed
10/23/2008 18:03:08|qmaster|etna|I|read job database with 22 entries in
0 seconds
10/23/2008 18:03:08|qmaster|etna|I|qmaster hard descriptor limit is set
to 8192
10/23/2008 18:03:08|qmaster|etna|I|qmaster soft descriptor limit is set
to 8192
10/23/2008 18:03:08|qmaster|etna|I|qmaster will use max. 8172 file
descriptors for communication
10/23/2008 18:03:08|qmaster|etna|I|qmaster will accept max. 99 dynamic
event clients

Above entries are from the moment I think it chrased. There are more
info some lines below:

10/24/2008 14:23:58|qmaster|xxxx|E|no connection open to berkeley
database "/aplicaciones/sge6.1u4//default/spool/spooldb"
10/24/2008 14:23:58|qmaster|xxxx|W|rule "default rule" in spooling
context "berkeleydb spooling" failed writing an object
10/24/2008 14:23:58|qmaster|xxxx|E|job 41074 was rejected cause it
couldn't be written
10/24/2008 14:23:58|qmaster|xxxx|E|Corrupted database detected. Freeing
all resources to prepare for a reconnect with recovery.
10/24/2008 14:23:58|qmaster|xxxx|E|error writing object with key "JOB:
41075" into berkeley database: (-30974) DB_RUNRECOVERY: Fatal error, run
database recovery

Any help is welcome

To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net

More information about the gridengine-users mailing list