On 02/20/2015 11:47 AM, Kenyon Ralph wrote:
>>> [pid  4797] write(2, "fatal out of memory (4048 bytes)"..., 33fatal out of 
>>> memory (4048 bytes)) = 33
>> Any ideas?
> Add to your ntp.conf:
> rlimit memlock 64

Thank you, that did not fix it, but point me in the right direction:
According to the docs on this command, that would set the limit of what
ntpd tries to lock even higher than the default of 32 MB, to 64 MB. The
default on SLES11 (and Ubuntu 14.04, and probably most others too) for
normal user processes however is 64 KILObytes, so anything higher than
"rlimit memlock 0" should crap out.
Setting "rlimit memlock 0" however did help.
As did setting the system ulimit to 131072 and removing the rlimit-entry
again. (I had actually tried to raise the ulimit, but not THAT high)

BTW: I think this new "feature" is setting new standards for braindeadness.
Using a default that will crap out on 90% of all systems out there is
not smart. As is locking such a huge amount of memory - on most systems
ntpd is not so important that you would want it to lock and thereby
permanently waste 32 MB of memory. Also, you could just as well call
getrlimit(RLIMIT_MEMLOCK to find out how much memory you actually are
allowed to use at runtime, and not exceed that - maybe printing out a
warning if there is a higher rlimit-setting in the configfile. Last but
not least: Practically all software that I know will simply use unlocked
memory if it is unable to lock memory, which would certainly be a better
alternative than dieing at some random point in time with a misleading
error message.
-- 
Michael Meier, Zentrale Systeme
Friedrich-Alexander-Universitaet Erlangen-Nuernberg
Regionales Rechenzentrum Erlangen
Martensstrasse 1, 91058 Erlangen, Germany
Tel.: +49 9131 85-28973, Fax: +49 9131 302941
[email protected]
www.rrze.fau.de
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool

Reply via email to