ctdb memory usage
ronniesahlberg at gmail.com
Thu Jan 7 15:55:44 MST 2010
Ok, i can make it a command line setting to control how big the log
area should be.
The purpose of this ringbuffer for logging is to be able to extract
useful log information from a system semi-immediately after an
incident has triggered.
(Since logging continously at level 10 to a log file may be unrealistic).
So it kind of allows you to log with a reasonable level to disk, but
once an abnormal incident has triggered, you can extract basically a
full level 10 log from memory.
This is MUCH nicer than "increase the log level to 10 and see if it
happens again" which is a suboptimal troubleshooting approach for
critical production systems. :-)
It would be _VERY_ nice to have something similar for samba.
On Fri, Jan 8, 2010 at 3:48 AM, Stefan (metze) Metzmacher
<metze at samba.org> wrote:
> Hi Ronnie,
> I noticed that versions > 1.0.108 use about 68MB ram,
> where 1.0.108 only used ~ 1MB.
> This is because of the large log ring buffer.
> This really a pain as my virtual machines start to swap badly.
> I think we should make this MAX_LOG_ENTRIES configureable.
> and change the default from 500000 to 10000 or so, which is
> ~ 4MB on my systems.
More information about the samba-technical