random ending of events

Post by **zoneminder** » Wed Oct 18, 2006 11:11 am

I have found a problem with zmaudit. I have done a fix and am testing it. It certainly won't hurt and should help remove bogusly deleted events. However I can't see any obvious connection between that and the event storm problem but it will certainly stop ZM getting into a tizzy when events are deleted that it is still trying to write to.

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 11:23 am

Well it would appear that it doesnt happen as regular as that system had been up for about 10 days before it started doing it. I know you said about not using a filter to restart but could i use a filter to send me an email when it happens so i can catch it early before it creates 28000 events!

Post by **zoneminder** » Wed Oct 18, 2006 11:42 am

Your system has gone doolally again, and this time the time checker is running. Here is an excerpt from the time check log which is getting the system time every half second or so.

Wed Oct 18 12:28:14 BST 2006
Wed Oct 18 12:28:15 BST 2006
Wed Oct 18 12:28:17 BST 2006
Wed Oct 18 12:28:18 BST 2006
Wed Oct 18 12:28:16 BST 2006
Wed Oct 18 12:28:17 BST 2006
Wed Oct 18 12:28:17 BST 2006
Wed Oct 18 12:28:18 BST 2006
Wed Oct 18 12:28:20 BST 2006
Wed Oct 18 12:28:21 BST 2006
Wed Oct 18 12:28:19 BST 2006
Wed Oct 18 12:28:22 BST 2006
Wed Oct 18 12:28:20 BST 2006
Wed Oct 18 12:28:23 BST 2006
Wed Oct 18 12:28:22 BST 2006
Wed Oct 18 12:28:22 BST 2006
Wed Oct 18 12:28:23 BST 2006
Wed Oct 18 12:28:25 BST 2006
Wed Oct 18 12:28:24 BST 2006
Wed Oct 18 12:28:24 BST 2006
Wed Oct 18 12:28:27 BST 2006

As you can see, the times are all over the place

I have never seen anything like this, and can't think what (apart from a horrendously corrupted ntp daemon) could cause this.

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 11:47 am

so do you think thats what it is and as the system tim eis jumping around is causing zm to do strage things?

It appears though noting else is affected but i suppose maybe zm is the only time critical process? It has done this accross many differnt kernels so im unsure as to what to do other than never use asus boards again

Post by **zoneminder** » Wed Oct 18, 2006 11:50 am

I have found this http://lkml.org/lkml/2003/3/26/32 which may be related, though I don't know what half of it means. But it does concern dual cpu AMD systems and that's what you've got!!

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 11:54 am

oh isee yes it is a dual proc amd. And if i recall this happened on my fx60 based test too but i assumed it was something in my distro!
Do you think if i use a single core (ie non smp kernel) it will go away?

Post by **zoneminder** » Wed Oct 18, 2006 12:06 pm

I think so, as it would possibly appear from that post to be a consequence of the two cpus getting out of sync. However I claim no particular knowledge of the issue.

Certainly times jumping around will knacker ZM, as it will think events have already ended before they have started etc and other horrible stuff. I would think it would probably screw up a lot of applications as well, but your system wasn't really doing anything else anyway.

The good news at least is that we have positively identified the issue, and I found a bug in zmaudit

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 12:08 pm

ok then i will drop to single core and cross my fingers!
You would think this would be a major issue though as the link you posted was from 2003. Looks like im defecting back to intel then!

Post by **zoneminder** » Wed Oct 18, 2006 12:25 pm

Well, it's possible that post has nothing to do with it. But it's incontrovertible that your system has an unreliable clock and that is pretty fatal. What distro did it have on it? Maybe it's worth trying a different one?

There was a mention in another post on the same site that the time stayed the same until you got something calling the system function gettimeofday a lot. Well ZM probably calls it a bunch, but maybe other applications are not so clock intensive and so the problem never gets triggered.

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 12:30 pm

it was on a pclinux install now its on ross's 2006 non live distro. But I also tried coreys distro

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 12:41 pm

im not sure i can leave it single core as its load is now
9.79 7.50 5.06

but i am running a filter to remove all events less than 5 seconds
and this place has a version of zm4ms running on quite high refresh rates

Post by **zoneminder** » Wed Oct 18, 2006 1:02 pm

The other post mentioned a boot parameter that might affect something or other to do with syncing the procs. Do you know if you can use that?

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 1:07 pm

If I understood what it meant then maybe?!?

Post by **zoneminder** » Wed Oct 18, 2006 1:23 pm

You should be able to interrupt your bootloader and add the notsc option to the boot line to temporarily experiment with. I don't know what other consequences it may have though.

jameswilson · Post by **jameswilson** » Wed Oct 18, 2006 1:27 pm

its one of those headless things can i add it to the kernel parameters in lilo, like i do for video drivers eg xdrv=vesa or reboot=bios ?

i assume there will be a variable too like sync=notsc or notsc=true