If anybody else happens to experience this issue, Daniel and I were
able to determine that it was due to a rather large fts-queue file
(4.4M in my case).  Removing the fts-queue file and letting OSSEC
create a new file allowed ossec-analysisd and ossec-logtest to start
instantly without excessive CPU usage.  Daniel is going to work on
improving the code that reads the fts-queue file.

Regards,
Doug Burks
http://securityonion.blogspot.com/

On Tue, Mar 9, 2010 at 2:41 PM, Doug Burks <[email protected]> wrote:
> Hi Daniel,
>
> Thanks for your response.  We're running OSSEC 2.3 on CentOS 5.4.
> Nothing unusual in ossec.conf or local_rules.xml (I sent you a direct
> email with a copy of my local rules).  We have 33 agents total (24
> Windows, 9 Linux).  All agents are running 2.3 as well.  ossec-logtest
> is exhibiting the same behavior; would it be affected by agents?  Is
> there any additional logging that I can enable to determine what is
> taking so much time and CPU?
>
> Thanks,
> Doug Burks
>
> On Mar 9, 7:41 am, Daniel Cid <[email protected]> wrote:
>> Hi Doug,
>>
>> I have no clue to what might be going on... syscheckd taking long
>> doesn't matter,
>> because it "sleeps"  in the middle to save some CPU. All normal..
>>
>> For analysisd and log-test to take that long, there must be something in your
>> rules or environment that's causing all that delay. I never had this
>> problem before...
>> What version are you using? Which OS? How many agents pointing to that box?
>>
>> Thanks,
>>
>> --
>> Daniel B. Cid
>> dcid ( at ) ossec.net
>>
>>
>>
>> On Fri, Mar 5, 2010 at 10:53 AM, Doug Burks <[email protected]> wrote:
>> > Yes, I saw that the log file showed a 3-minute gap between syscheckd
>> > starting and finishing pre-scan.  However, ossec-syscheckd is not the
>> > process that is taking up 100% CPU.  ossec-analysisd takes 100% CPU
>> > for 3 minutes.  ossec-logtest does the same thing, and I wouldn't
>> > expect it to do anything with syscheckd.
>>
>> > I've looked at 2 other OSSEC installs and neither of them exhibit this
>> > behavior.  When starting OSSEC, they do show the standard 3-minute
>> > syscheckd gap in the log file, but there is NO process taking 100% CPU
>> > for any amount of time.  Also, starting ossec-logtest on these other
>> > OSSEC installs is instantaneous with no excessive CPU usage.
>>
>> > What would cause ossec-analysisd and ossec-logtest to hit 100% CPU
>> > usage for 3 minutes?  Any ideas, Daniel Cid?
>>
>> > Thanks,
>> > Doug Burks
>>
>> > On Mar 4, 4:02 pm, Joshua Gimer <[email protected]> wrote:
>> >> On Thu, Mar 4, 2010 at 12:11 PM, Doug Burks <[email protected]> wrote:
>> >> > As I mentioned in my previous message, ossec-logtest takes about 3
>> >> > minutes before it will accept input.  During this time, it is stuck at
>> >> > 100% CPU usage.  ossec-analysisd does the same thing when starting
>> >> > OSSEC.  After the 3 minutes is up, ossec-analysisd settles down to
>> >> > about 30% CPU usage.
>>
>> >> > ....
>> >> > 2010/03/04 13:59:55 ossec-syscheckd: INFO: Starting syscheck database
>> >> > (pre-scan).
>> >> > 2010/03/04 14:02:41 ossec-syscheckd: INFO: Finished creating syscheck
>> >> > database (pre-scan completed).
>>
>> >> > Is this normal?
>>
>> >> > Thanks,
>> >> > Doug Burks
>>
>> >> The majority of the time is being spent starting the syscheck database.
>> >> Google seems to have a few results of OSSEC start logs that show around a 
>> >> 3
>> >> minute start as well.
>>
>> >> --
>> >> Thx
>> >> Joshua Gimer
>



-- 
Doug Burks, GCIA, GSEC, CISSP
http://securityonion.blogspot.com

Reply via email to