eDirectory Log Rotation (ndsd.log)

[Anonymous_User]

Hi all,

We're being plagued by a bug which fills up our ndsd.log and makes it
grow really large. I presume that people working with idm 4.01 and java
1.6 have the same issues...
We're preparing to update the systems, but in the meanwhile, we need to
rotate the log, which didn't go well last night.

Following the knowledgebase
(http://www.novell.com/support/kb/doc.php?id=7005454 , which has now
disappeared) article, I tried to do the same on RH 6.3. This does
however only partly work, because ndsd and lcache keep a grip on the old
log file and keeps filling it, while it is deleted. Jolly!

So I guess ab was right in his answer in post
https://forums.netiq.com/archive/index.php/t-44434.html , that kilall
-HUP ndsd is not really working the way it should.

The question is, has anyone (since the post dates back to 2010) find
another solution for the logrotate issue? Or are there alternatives
besides monitoring the log size and restarting the processes manually?

Kind regards,
Tom


--
dedobbeleer
------------------------------------------------------------------------
dedobbeleer's Profile: https://forums.netiq.com/member.php?userid=4764
View this thread: https://forums.netiq.com/showthread.php?t=48259

[Anonymous_User]

Simple way to clean up a log file, even while handles from other processes
are watching it:

Code:
----------
echo -n '' > /path/to/ndsd.log
----------

Of course, this isn't log rotation, but the ndsd.log is not usually that
important. You could always copy out the data before doing this if you
really wanted:

Code:
----------
cat /path/to/ndsd.log | bzip2 > /path/to/ndsd-`date +%s`.log.bz2
----------

Good luck.

[Anonymous_User]

On Wed, 24 Jul 2013 08:14:01 +0000, dedobbeleer wrote:

> We're being plagued by a bug which fills up our ndsd.log and makes it
> grow really large. I presume that people working with idm 4.01 and java
> 1.6 have the same issues...

Details? I'm not seeing anything like this here, but I'm on IDM402.


> We're preparing to update the systems, but in the meanwhile, we need to
> rotate the log, which didn't go well last night.

How are you attempting to do this?


> Following the knowledgebase
> (http://www.novell.com/support/kb/doc.php?id=7005454 , which has now
> disappeared) article, I tried to do the same on RH 6.3.

Hm. Ok, so where did you find it, if it's not there any more. I'll see if
I can find out where it went.


> This does
> however only partly work, because ndsd and lcache keep a grip on the old
> log file and keeps filling it, while it is deleted. Jolly!
>
> So I guess ab was right in his answer in post
> https://forums.netiq.com/archive/index.php/t-44434.html , that kilall
> -HUP ndsd is not really working the way it should.

As he said, there's no reason to think sending HUP to ndsd is going to do
anything useful at all. Maybe it did once, for somebody, but I've not
seen it documented anywhere that I'd trust.


> The question is, has anyone (since the post dates back to 2010) find
> another solution for the logrotate issue? Or are there alternatives
> besides monitoring the log size and restarting the processes manually?

If you must, then I don't see why this wouldn't work:

/etc/init.d/ndsd stop
.... do whatever to ndsd.log ...
/etc/init.d/ndsd start


--
--------------------------------------------------------------------------
David Gersic dgersic_@_niu.edu
Knowledge Partner http://forums.netiq.com

Please post questions in the forums. No support provided via email.

[Anonymous_User]

Hi all,

ab, thanks for the suggestions, I'll look into it.

David, answering your questions:
1) The bug is regarding the "unloading class" messages. Apparently, this
can be fixed with java 1.7.
2) We're rotating it with a simple logrotate config. I think you can
find it in the other forum post as well. Nothing special, except for the
hangup i mentioned. It was in that knowledgebase article, so I hoped...
3) I must, but I'd rather don't 🙂 I don't want our engineers or myself
to getting up at 3am for a failed ndsd restart 🙂 I'd rather play safe.

Thanks for the feedback!
Tom


--
dedobbeleer
------------------------------------------------------------------------
dedobbeleer's Profile: https://forums.netiq.com/member.php?userid=4764
View this thread: https://forums.netiq.com/showthread.php?t=48259

[Anonymous_User]

I think I found the answer: copytruncate in the logrotate conf. That
copies the ndsd.log and truncates it afterwards.
Simple as that! Will test it first though 🙂


--
dedobbeleer
------------------------------------------------------------------------
dedobbeleer's Profile: https://forums.netiq.com/member.php?userid=4764
View this thread: https://forums.netiq.com/showthread.php?t=48259

[Anonymous_User]

On Wed, 24 Jul 2013 14:54:01 +0000, dedobbeleer wrote:

> 1) The bug is regarding the "unloading class" messages. Apparently, this
> can be fixed with java 1.7.

Sorry, I haven't seen this. Is there an actual bug here?


> 2) We're rotating it with a simple logrotate config. I think you can
> find it in the other forum post as well. Nothing special, except for the
> hangup i mentioned. It was in that knowledgebase article, so I hoped...

The TID you referenced has been pulled deliberately as being unsupported.
As you found, it doesn't work. Given that, I'd go with "don't do that" as
the best possible recommendation.


> 3) I must, but I'd rather don't 🙂 I don't want our engineers or myself
> to getting up at 3am for a failed ndsd restart 🙂 I'd rather play safe.

I guess "safe" is a matter of opinion. If you're not using the contents
of ndsd.log anyway, "ln -s /dev/null ndsd.log" would be "safe", too.


--
--------------------------------------------------------------------------
David Gersic dgersic_@_niu.edu
Knowledge Partner http://forums.netiq.com

Please post questions in the forums. No support provided via email.