In all the time I've traced the code - and I've literally spent months on this on and off - there is nothing that would cause any data loss.
AFAIK, --disable-mutexextrainfo is perfectly safe. That's why I put it in there... The problem is really a basic limitation of POSIX. gcc optimization (-O2) doesn't help either as it can execute things out of order and throw multiple CPUs in there and poof. We definitely want to keep -O2, so I guess we could put the mutex routines into a separate module, compile that ONE module w/o any optimization. But that's a lot of work and still doesn't completely 'fix' things. The problem is that there's no way to add information to a mutex, nor to recover information from the mutex (which is an opaque structure). So wtrying to track information about mutex usage (it's lock/unlock) as a way of tracing down deadlocks - is incredibly tricky. You are writing code to wrap around a POSIX mutex, which can't protect itself (or it single threads the entire app), etc. I've posted diatribes on this before - look in ntop-dev if you want more of the gore. -----Burton -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Arn Vollebregt Sent: Tuesday, January 24, 2006 4:44 AM To: [email protected] Subject: Re: [Ntop] queuePacket/dequeuePacket errors From: "Burton Strauss" <[EMAIL PROTECTED]> > You can use Gmane or Google to search the archives Ah, didnt know Gmane yet, usefull service. Not finding it via google was my own fault, my search query was to specific. > both are documented in docs/FAQ. Ok, I have now found your FAQ when noticing your email address. Perhaps you can convince the maintained of ntop.org to refer to your updated FAQ? Might safe some emails on the mailinglist ;) > Read the stuff in the archives and docs/FAQ on self-LOCKED - it will > give you a basis for the discussion. Okay, that gave me some idea on what these warnings are about. For now I added --disable-mutexextrainfo to my ntop.conf for I cant find yet on how to force ntop to a single CPU yet. That seems like a shame anyway, having 2 dual core CPU's and restricting ntop (one of the most resource hungy processes on the system) to a single CPU. I am still experiencing the slow web-interface at times, I guess it's 'collides' with a hash flush or some other update from ntop, trying a couple of seconds later it loads okay again. Thank you for your patience and time, and I'd like to say you did quite a nice job on the FAQ. Arn Vollebregt -- "Wisdom lies not in obtaining knowledge, but in using it in the right way" - kroesjnov > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf > Of Arn Vollebregt > Sent: Saturday, January 21, 2006 11:48 AM > To: [email protected] > Subject: Re: [Ntop] queuePacket/dequeuePacket errors > > From: "Burton Strauss" <[EMAIL PROTECTED]> >> Read the back traffic on ntop - it's a multiple threaded app. > > My bad, didnt know this. > On reading back; I just joined the list, and the webarchives dont > provide a search function, so 'reading back' isnt very feasable imho. > >> Self-locked really shouldn't happen in 3.2, although it's just a >> warning - > > Well, apparently it does happen :) > What does this self-locked actually mean for me? Will I loose the data > concerned? Will it be queued? > And more important; Are there any pointers on what is causing these > errors? > >> usually the data at risk is only the locking data itself. > > If it would occur a couple of times a day I would not be very > concerned, but my logs are really filling up with these errors. I > guess there is a decent amount of data at risk > >> As a test, you can try using >> the OS to restrict ntop to one processor (if it will allow you to do so). > > I'll look into this on monday when I have access to the server again. > > NTOP is also having peaks of 90% CPU load, this also doesnt seem very > normal to me? > tia for your time. > > Arn Vollebregt > > -- > "Wisdom lies not in obtaining knowledge, but in using it in the right way" > - kroesjnov > >> -----Original Message----- >> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf >> Of Arn Vollebregt >> Sent: Friday, January 20, 2006 4:24 AM >> To: [email protected] >> Subject: [Ntop] queuePacket/dequeuePacket errors >> >> Hello, >> >> I am having some troubles with running NTOP on a 100Mbit network with >> quite some traffic to monitor. A short while after starting up the >> process the webinterface becomes sluggisch and unresponsive at times. >> When I check /var/log/messages I am seeing a flood of the errors >> below repeating >> >> **WARNING** tryLockMutex() called 'queuePacket' with a self-LOCKED >> mutex [EMAIL PROTECTED]:2219] >> **WARNING** accessMutex() called 'dequeuePacket' with a self-LOCKED >> mutex [EMAIL PROTECTED]:2399] >> >> NTOP is also spawning multiple processes, at the time of writing I am >> up to >> 9 NTOP processes, running for a small 5 minutes now. >> >> I recon NTOP cannot keep up with the traffic or the amount of hosts >> and their data, are there ways to tweak NTOP? I figure the machine >> itself is not the problem, since it concerns a dual Xeon 2.4Ghz, 1Gig >> of RAM with a SCSI harddisk. >> >> Arn Vollebregt >> >> -- >> "Wisdom lies not in obtaining knowledge, but in using it in the right >> way" >> - kroesjnov >> >> _______________________________________________ >> Ntop mailing list >> [email protected] >> http://listgateway.unipi.it/mailman/listinfo/ntop >> >> _______________________________________________ >> Ntop mailing list >> [email protected] >> http://listgateway.unipi.it/mailman/listinfo/ntop >> >> >> >> -- >> No virus found in this incoming message. >> Checked by AVG Free Edition. >> Version: 7.1.375 / Virus Database: 267.14.21/235 - Release Date: >> 19-1-2006 >> > > _______________________________________________ > Ntop mailing list > [email protected] > http://listgateway.unipi.it/mailman/listinfo/ntop > > _______________________________________________ > Ntop mailing list > [email protected] > http://listgateway.unipi.it/mailman/listinfo/ntop > > > > -- > No virus found in this incoming message. > Checked by AVG Free Edition. > Version: 7.1.375 / Virus Database: 267.14.21/236 - Release Date: > 20-1-2006 > _______________________________________________ Ntop mailing list [email protected] http://listgateway.unipi.it/mailman/listinfo/ntop _______________________________________________ Ntop mailing list [email protected] http://listgateway.unipi.it/mailman/listinfo/ntop
