Re: [Bacula-users] File selection via bconsole takes a long timewhen 1, 4 million of files to be selected for restore

[Jürgen K. <Jue...@we...>]

Hello John,

> One of the big causes of having a slow build of the file list is that
> your postgresql database is not properly indexed. Ralf asked about
> disk IO and you said there was not much so I doubt this is the cause.
> If you have an improperly indexed database there will be constant disk
> activity.

I understood this. Assuming we have an improperly indexed database and the involved tables and the corresponding indexes fit into the servers memory we will not encounter much I/O activity but more CPU consumption of the pgsql-client, but this not the case here. Instead of this the bacula-dir process is consuming a lot of cpu resources. I tried to strace the bacula-dir process trying to find out what's going on, but the bacula-dir process does not like this, it seems to be blocked in the end and you cannot abort strace with a SIGINT you must send him a SIGKILL to abort strace and restart the bacula-dir process for properly function. I started the bacula-dir process in debug mode (e.g. bacula-dir -f -d200) but  - to me - it did not report a comprehensible explanation. The next reasonable step could be, with the help of the debug messages to go into the source code of bacula-dir and try to find an explanation - but this could be time-consuming and besides of this I am not very familiar with C/C++.

Thank you John and Ralf,
Jürgen