cacti 0.8.6d: polling process lock up

hoganw · Post by **hoganw** » Sat May 07, 2005 2:55 pm

Downloaded the new poller.php, now it doesn't run cactid at all.

I get Content-type: and X-Powered-By: and then it returns me to my shell prompt, no errors, nada.

Ideas?

[EDIT]

If I manually call cmd.php it runs but is only running one poller at a time.
Same if I manually run cactid. So, for some reason the new poller.php isn't running the specified poller?

Post by **TheWitness** » Sat May 07, 2005 8:31 pm

You likely overwrote <path_cacti>/poller.php with <path_cacti>/lib/poller.php. Not good.

TheWitness

hoganw · Post by **hoganw** » Sat May 07, 2005 11:05 pm

TheWitness wrote:You likely overwrote <path_cacti>/poller.php with <path_cacti>/lib/poller.php. Not good.

TheWitness

Ack, and I'm usually so good about catching that sort of thing.

I'll post an update tomorrow and let you know if that fixed my gaps. Thanks and sorry for being a dummy.

hoganw · Post by **hoganw** » Sun May 08, 2005 1:45 pm

No change with the new lib/poller.php, still getting the same gaps.

Is there any way to roll back to 'c' short of dropping the db, loading the backup and putting the old version back? At this point a working cacti is more important then having the newest cacti since we do use it for watching production servers and trending.

If there's something else I can try before rolling back, let me know, otherwise I really have no choice.

Thanks for the help so far!

Post by **TheWitness** » Sun May 08, 2005 1:47 pm

You can replace the previous files (0.8.6c) with patches and then change the version table to 0.8.6c and you will be finished.

TheWitness

Did you update any other software? PHP, MySQL, Net-SNMP, etc. Are you using Cactid or cmd.php?

TheWitness

Post by **TheWitness** » Sun May 08, 2005 1:48 pm

Can you AIM or MSM?

TheWitness

Post by **TheWitness** » Sun May 08, 2005 1:49 pm

Or NetMeeting?

TheWitness

Post by **TheWitness** » Sun May 08, 2005 8:15 pm

Please try this cmd.php and let me know if it corrects the problem.

TheWitness

hoganw · Post by **hoganw** » Mon May 09, 2005 3:47 am

I'm using cactid and I'm on MSN/AOL/Yahoo (prefer Yahoo). So far I haven't changed back, so if you'd like me to try cmd.php instead of cactid that's not a problem. PM me your IM info and I'll get in touch with you tomorrow (Monday) if you have time? I'll be at work and be able to dedicate time to this pretty much any time between 9am and 5pm PST (GMT-8). Right now I'm going to bed.

Thanks again

Post by **TheWitness** » Mon May 09, 2005 4:11 am

I am suspecting that the problem, if not due to the lib/poller changes is due to an issue with snmp timeouts getting the sysUptime variable from heavily loaded systems. Therefore, I have changed the behavior of cmd.php to not issue a spike kill when it does not receive a valid sysUptime response. Before you go increasing your timeouts, try this out first. If it works, I will move into Cactid 0.8.6e RC6 as well.

TheWitness

hoganw · Post by **hoganw** » Mon May 09, 2005 11:29 am

Ok, I've put the new cmd.php into place. I'll let you know if there have been any changes.

As far as heavily loaded servers go, the cacti box is the one getting beat up.

I have it set for 5 poller processes and 5 threads per process, box gets reeeeeally slow for about 30 seconds.

hoganw · Post by **hoganw** » Mon May 09, 2005 3:53 pm

Lots of things happened today!

My new cacti server finally got built and is currently being tested and tweaked. New server is Linux x86_64 (AMD64). Got the .rrd files converted for the new arch, db data copied over, etc. In the process I went back to 0.8.6c as well.

Just wanted to let you know where I was at today and that certain things have or will be changing soon.

Lojak · Post by **Lojak** » Sat May 14, 2005 4:54 pm

I'm experiencing a similar problem. I've used the new cmd.php and poller.php (replacing the poller.php file in lib/), and some graphs have been generated, but others have not.

Number of hosts: 65
Number of scripts: 5
RRD files: 500

OS: FreeBSD 5.3
MySQL: 4.1.11
PHP: 4.3.10
Net-SNMP: 5.2.1

When I use cmd.php, these is the result I'm now getting:

05/14/2005 05:50:01 PM - PHPSVR: Poller[0] PHP Script Server has Started - Parent is cmd
05/14/2005 05:50:01 PM - CMDPHP: Poller[0] PHP Script Server Started Properly
05/14/2005 05:50:01 PM - PHPSVR: Poller[0] ERROR: Input Expected, Script Server Terminating

Manually running php poller.php outputs this:

Waiting on 1/1 pollers.
Waiting on 1/1 pollers.
Waiting on 1/1 pollers.
Waiting on 1/1 pollers.
[...]

BUT occassionally it works, and the information is polled and the information is graphed properly.

Very weird.

Unfortunately I've fubared by back-up of my 0.8.6c DB. Is it possible to down-grade to that version? I'll make the changes to the DB manually if I have to.

-- Stephen

Post by **TheWitness** » Sat May 14, 2005 8:09 pm

Stephen,

To backoff the 0.8.6d database, just replace the files in the various directories and modify the version table in mysql. Change the value in the one and only row to 0.8.6c

TheWitness

p.s. I believe this error is due to a problem in php. You might just be able to back off the lib/poller.php from the previous version. Give that a whirl first .

Lojak · Post by **Lojak** » Sat May 14, 2005 8:32 pm

Thanks very much for the info (my boss will be happy now). I tried rolling back the lib/poller.php and I had no success - lots of "Waiting on 1/1 pollers" type messages, with no apparent updating.

Thanks for you help

-- Stephen.

cacti 0.8.6d: polling process lock up

Who is online