2.3.15, murder, skiplist aborts

[John Madden]

Looking through the archives it looks like Bron produced a patch that 
puts this abort in to prevent corruption (yay) and instead print aborts, 
but I now have a situation where things appear to be hung and I'm trying 
to find out why and what I should do to fix it.

(This all happened while batch-creating about 300k mailboxes but I only 
got a couple thousand in before the borking.)

The master and backends appear to be ok, but the frontend, where the 
creates are going, sees this:

Oct 27 13:43:34 imap ctl_cyrusdb[20242]: checkpointing cyrus databases
Oct 27 13:43:34 imap ctl_cyrusdb[20242]: done checkpointing cyrus databases
Oct 27 13:44:00 imap mupdate[3134]: skiplist: checkpointed 
/var/imap/mailboxes.db (15607 records, 1183084 bytes) in 0 seconds
Oct 27 13:44:12 imap imap[20158]: Fatal error: Internal error: assertion 
failed: cyrusdb_skiplist.c: 727: db->current_txn == NULL
Oct 27 13:44:12 imap imap[20158]: skiplist: closed while still locked
Oct 27 13:46:42 imap imap[20323]: auxpropfunc error invalid parameter 
supplied
Oct 27 13:46:47 imap imap[20205]: login: localhost.localdomain 
[127.0.0.1] cyrus PLAIN User logged in
Oct 27 13:47:26 imap mupdate[3134]: skiplist: checkpointed 
/var/imap/mailboxes.db (19188 records, 1454316 bytes) in 1 second
Oct 27 13:48:34 imap ctl_cyrusdb[20332]: checkpointing cyrus databases
Oct 27 13:48:34 imap ctl_cyrusdb[20332]: done checkpointing cyrus databases
Oct 27 13:49:31 imap imap[18754]: skiplist: closed while still locked
Oct 27 13:51:11 imap imap[20206]: skiplist: closed while still locked
Oct 27 13:53:08 imap imap[18753]: skiplist: closed while still locked
Oct 27 13:53:34 imap ctl_cyrusdb[20381]: checkpointing cyrus databases
Oct 27 13:53:34 imap ctl_cyrusdb[20381]: done checkpointing cyrus databases
Oct 27 13:54:37 imap imap[18758]: skiplist: closed while still locked

Does any of that need to be worried about?  While running (a 
multi-threaded "go make a ton of mailboxes as quickly as you can,") I 
noticed that the connection some of the backends stops.  If it's just 
that the imapd generating the abort is the one responsible, that's fine, 
but how do we prevent this situation to begin with?

John




-- 
John Madden
Sr UNIX Systems Engineer
Ivy Tech Community College of Indiana
jmadden at ivytech.edu