Putting queue into tmpfs helped a lot, now I'm thinking about storing it permanently and dumping with cronjob regularly.
Btw, I'm using spool instead of kannel-store currently. Is there any performance difference between spool and store (when queue is in 1 file) ? 2012/10/16 spameden <[email protected]> > Is it possible to have store_load in a separate thread? So SQLBox / SMSBox > would connect before store fully loaded? > > > 2012/10/16 spameden <[email protected]> > >> Hi. >> >> I've copied both devel and users list to get quicker response, sorry if >> it's not right. >> >> We are suffering from quite slow store load (about 100sms/sec) with large >> queue (about 100k MTs). >> >> What could be a possible reason? Could it be slow disk or is there a bug >> itself inside the kannel? >> >> >> Also we encountered another error while restarting bearerbox: >> 2012-10-16 12:26:02 [9976] [9] PANIC: gwlib/gw-rwlock.c:137: >> gw_rwlock_rdlock: Assertion `lock != NULL' failed. >> 2012-10-16 12:26:02 [9976] [9] PANIC: >> /usr/sbin/bearerbox(gw_backtrace+0xae) [0x48ba6e] >> 2012-10-16 12:26:02 [9976] [9] PANIC: /usr/sbin/bearerbox(gw_panic+0x145) >> [0x48bbc5] >> 2012-10-16 12:26:02 [9976] [9] PANIC: >> /usr/sbin/bearerbox(gw_rwlock_rdlock+0x54) [0x4818e4] >> 2012-10-16 12:26:02 [9976] [9] PANIC: >> /usr/sbin/bearerbox(route_incoming_to_boxc+0x4b) [0x4126db] >> 2012-10-16 12:26:02 [9976] [9] PANIC: >> /usr/sbin/bearerbox(bb_smscconn_receive+0x628) [0x419728] >> 2012-10-16 12:26:02 [9976] [9] PANIC: >> /usr/sbin/bearerbox(bb_smscconn_sent+0x110) [0x41a3d0] >> 2012-10-16 12:26:02 [9976] [9] PANIC: /usr/sbin/bearerbox() [0x4596af] >> 2012-10-16 12:26:02 [9976] [9] PANIC: /usr/sbin/bearerbox() [0x45af97] >> 2012-10-16 12:26:02 [9976] [9] PANIC: /usr/sbin/bearerbox() [0x48292e] >> 2012-10-16 12:26:02 [9976] [9] PANIC: /lib/libpthread.so.0(+0x68ca) >> [0x7fcd35d1a8ca] >> 2012-10-16 12:26:02 [9976] [9] PANIC: /lib/libc.so.6(clone+0x6d) >> [0x7fcd3538fb6d] >> >> NOTE: we are using simple scheme bearerbox <-> sqlbox and when we are >> restarting we are closing sqlbox prior to bearerbox >> > >
