PubSub agent and networked home dirs
This isn't exactly OS X server related, but the brain trust to answer my question is more likely to be in this forum than others. Quick background: I serve home directories off a ReadyNAS via netatalk, and manage an openldap directory via Workgroup Manager.
Some of my users are having odd problems which I think started with 10.6.2 or 10.6.3. If a client host goes to sleep, on wakeup there is an afp reconnect issue that eventually works itself out. However, the PubSubAgent becomes quite angry. Example:
4/13/10 10:28:20 PM kernel AFP_VFS afpfs_DoReconnect: Restoring session /Network/Servers/readynas.local/Users
4/13/10 10:28:20 PM KernelEventAgent[43] tid 00000000 received event(s) VQ_NOTRESP (1)
4/13/10 10:28:20 PM PubSubAgent[15149] SQL Error: SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"
4/13/10 10:28:20 PM PubSubAgent[15149] SQLite::Exception "SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"" caught in void Foundation::AbstractTimer::_fire()
4/13/10 10:28:21 PM kernel AFP_VFS afpfs_DoReconnect: Primary Reconnect failed 5 on /Network/Servers/readynas.local/Users
4/13/10 10:28:21 PM kernel AFP_VFS afpfs_DoReconnect: trying Secondary Reconnect on /Network/Servers/readynas.local/Users
4/13/10 10:28:21 PM kernel AFP_VFS afpfsCheckOpenFiles: File <Database.sqlite3> can not be reopened due to deny modes or byte range locks.
4/13/10 10:28:21 PM kernel AFP_VFS afpfsCheckOpenFiles: File <WebpageIcons.db> can not be reopened due to deny modes or byte range locks.
4/13/10 10:28:21 PM kernel ASP_TCP CancelOneRequest: cancelling slot 23 error 35 reqID 61046 flags 0x29 afpCmd 0x22 so 0x7e02000
4/13/10 10:28:21 PM kernel ASP_TCP CancelOneRequest: cancelling slot 24 error 35 reqID 61047 flags 0x29 afpCmd 0x7A so 0x7e02000
4/13/10 10:28:21 PM kernel AFP_VFS afpfs_DoReconnect: get the reconnect token
4/13/10 10:28:23 PM PubSubAgent[15149] SQL Error: SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"
4/13/10 10:28:23 PM PubSubAgent[15149] SQLite::Exception "SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"" caught in void Foundation::AbstractTimer::_fire()
The PubSubAgent's sqlite errors repeat every 3 seconds, during which time safari is crippled and takes forever to start, then periodically freezes. I verified that the sqlite3 db is fine. I can even run the queries that are failing. If I delete the db, then the problem occurs again after a sleep/wake cycle. It appears to have something to do with afpfsCheckOpenFiles losing a lock on the db.
Anyone have advice?
Some of my users are having odd problems which I think started with 10.6.2 or 10.6.3. If a client host goes to sleep, on wakeup there is an afp reconnect issue that eventually works itself out. However, the PubSubAgent becomes quite angry. Example:
4/13/10 10:28:20 PM kernel AFP_VFS afpfs_DoReconnect: Restoring session /Network/Servers/readynas.local/Users
4/13/10 10:28:20 PM KernelEventAgent[43] tid 00000000 received event(s) VQ_NOTRESP (1)
4/13/10 10:28:20 PM PubSubAgent[15149] SQL Error: SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"
4/13/10 10:28:20 PM PubSubAgent[15149] SQLite::Exception "SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"" caught in void Foundation::AbstractTimer::_fire()
4/13/10 10:28:21 PM kernel AFP_VFS afpfs_DoReconnect: Primary Reconnect failed 5 on /Network/Servers/readynas.local/Users
4/13/10 10:28:21 PM kernel AFP_VFS afpfs_DoReconnect: trying Secondary Reconnect on /Network/Servers/readynas.local/Users
4/13/10 10:28:21 PM kernel AFP_VFS afpfsCheckOpenFiles: File <Database.sqlite3> can not be reopened due to deny modes or byte range locks.
4/13/10 10:28:21 PM kernel AFP_VFS afpfsCheckOpenFiles: File <WebpageIcons.db> can not be reopened due to deny modes or byte range locks.
4/13/10 10:28:21 PM kernel ASP_TCP CancelOneRequest: cancelling slot 23 error 35 reqID 61046 flags 0x29 afpCmd 0x22 so 0x7e02000
4/13/10 10:28:21 PM kernel ASP_TCP CancelOneRequest: cancelling slot 24 error 35 reqID 61047 flags 0x29 afpCmd 0x7A so 0x7e02000
4/13/10 10:28:21 PM kernel AFP_VFS afpfs_DoReconnect: get the reconnect token
4/13/10 10:28:23 PM PubSubAgent[15149] SQL Error: SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"
4/13/10 10:28:23 PM PubSubAgent[15149] SQLite::Exception "SQLITE_IOERR[10.15]: disk I/O error in "select feed from Subscriptions where refreshInterval>=0 and feed>0 group by feed"" caught in void Foundation::AbstractTimer::_fire()
The PubSubAgent's sqlite errors repeat every 3 seconds, during which time safari is crippled and takes forever to start, then periodically freezes. I verified that the sqlite3 db is fine. I can even run the queries that are failing. If I delete the db, then the problem occurs again after a sleep/wake cycle. It appears to have something to do with afpfsCheckOpenFiles losing a lock on the db.
Anyone have advice?
Core Duo 2.16 MBP, Mac OS X (10.6.3), 2g iPhone;C2D Mini;3 x g4 Mini;AE 802.11n;AE dualband 802.11n;3 x AX 802.11n
