[ejabberd] muc_online_room crashing cluster?

Stephan Maihöfer sma at turtle-entertainment.de
Tue Jan 12 14:31:44 MSK 2010


we are encountering severe problems with our ejabberd-cluster
setup at the moment. we are seeing that an ejabberd-node
crashes once it reaches 15.000 elements in the muc_online_room
table. that table is set to ram&disc copy and has about
a gigabyte in size at that moment, while each node has
four gigabyte of ram and thus should be abled to handle this...even
when holding all the data in RAM. in other monitoring scripts
we don't see that the nodes run agains a RAM limit as well,
even the load is quite ok.

the cluster basically consists of six nodes, four of which
are handling connections from the outside, two of these just
hold the database in case other nodes go down.

any idea on how this can happen and/or what we do wrong?

thanks in advance,
stephan maihöfer

More information about the ejabberd mailing list