On Tue, Aug 03, 1999 at 09:05:04AM -0400, mr@noc.1st.net wrote:
Here's what it looked like today:
root 13410 0.3 54.2 70452 69716 ? S Aug02 5:00 syslog-ng
I've attached the conf file.
What kind of system do you have? I mean is it Linux or something else? I tried to track down that memory leak, but syslog-ng was not growing at all for any of the stresstests I used. (I tried both the file and the udp source driver, and file destination driver) I see the following possibilities: 1) your malloc() implementation in the libc doesn't like fragmentation. syslog-ng allocates lots of memory blocks, but frees them as well, maybe your malloc() implementation may have problems with fragmented memory. You may have a look at GNU Malloc, which is AFAIR available separate from libc, and can be compiled to a standalone shared lib. 2) The output queues of syslog-ng begin to fill up, because it is unable to handle the load. This should not be a problem though, since by default the output queues are 100 entries long for each file. (using 200 bytes as an estimated length of a message we get 20*100*200 = 400000 Bytes for twenty file destinations) 3) The libc itself leaks somewhere. I use GNU libc 2.1.1, GNU/Linux Debian, potato distribution. -- Bazsi PGP info: KeyID 9AF8D0A9 Fingerprint CD27 CFB0 802C 0944 9CFD 804E C82C 8EB1 url: http://www.balabit.hu/pgpkey.txt
On Wed, 4 Aug 1999, Balazs Scheidler wrote (abridged):
On Tue, Aug 03, 1999 at 09:05:04AM -0400, mr@noc.1st.net wrote:
Here's what it looked like today:
root 13410 0.3 54.2 70452 69716 ? S Aug02 5:00 syslog-ng
I've attached the conf file.
What kind of system do you have?
The machine is a Pentium III 450 running RedHat Linux 6.0 (which also uses GNU libc 2.1.1). It seems to be functioning properly in all respects except for the fact that the process continuously grows. All of the information it logs is up to date [not delayed] leading me to believe that the queues are not being filled. Regards, Jonathan Chapman
participants (2)
-
Balazs Scheidler
-
mr@noc.1st.net