-
Because of an omission when copying cur_block fields to new_output while processing a "logdir" config line, the "postrotate_cmd" option has no effect if it occurs before "logdir" in a block.
This is particularly a problem because it means specifying a "postrotate_cmd" in the default options section, where there is not normally a "logdir" option at all, has no effect.
This simple addition...
2009-10-27 10:58:30 UTC by mwhitlock
-
I think this prolect is dead. I may be wrong, but AFAIK there is no maintainer.
2009-08-11 08:32:00 UTC by low
-
I try to log several sections into one log directory, so all stuff concerning wireless should go into /var/log/wireless:
b43 :
facility = "kern"
regex = "b43-phy0"
logdir = "/var/log/wireless"
break = 1
wpa_cli :
regex = "wpa_cli"
logdir = "/var/log/wireless"
break = 1
This seems to be not...
2009-08-11 08:09:37 UTC by massimo_b
-
Hi. Is there still some work in progress on this project?
About my old question I'm using now
command = "echo"
to suppress any kind of messages.
2009-08-11 07:59:34 UTC by massimo_b
-
The signal handlers that are set up for the parent gets propagated to the child kernel logger process after fork(). The handlers should be reset to default. Otherwise among other issues, when the child tries to write to the uninitialized socket pair, the write will fail. Please also note that only the parent initializes the signal queueing socket pairs.
Suggesting to add the following patch...
2009-07-03 16:39:31 UTC by anisinha
-
Using --async should have pretty much the same performance effect, but logs would still end up on disk eventually, unless the computer crashes before flushing its logs.
In recent metalog versions, there is a flush option for individual logs, so you can have some logs flushed after each message, and others not. sysklogd provides a similar feature.
2009-06-11 09:26:26 UTC by gagern
-
Unfortunately it seems metalog still can't read something other than /dev/log.
Due to this issue I was forced to switch to syslog-ng on my server.
It's a so pity that lack of this small feature make this nice logger absolutely unusable.
2009-05-24 11:32:24 UTC by bircoph
-
My first implementation was flawed. In scenarios with rates slightly below limit for long periods of time, the unused rate wouldn't accumulate to be used in a burst. I finally recalled the concept of a token bucket, with one token being added every tick, and each message taking a token and being printed, or encountering an empty bucket and being discarded. This algorithm is more correct, more...
2009-04-21 16:41:24 UTC by gagern
-
I wrote a patch to limit the number of messages per unit of time that metalog will write to the log files.
My guiding light here was the limit module for iptables, and though I didn't look at its implementation, I used a very similar configuration with a "limit" option giving e.g. "5/minute" for a maximum of five messages per minute and a "limit_burst" option giving the maximum number of...
2009-04-21 09:26:03 UTC by gagern
-
vapier changed the public information on the Metalog project.
2009-03-16 07:44:36 UTC by vapier
