Gaps in Graphs since upgrading to v0.8.6b on FC2

_JPL_ · Post by **_JPL_** » Tue Nov 30, 2004 12:13 pm

I've been noticing periodic gaps in my graphs (all graphs) since upgrading to version 0.8.6b from 0.8.5 of cacti using poller.php. However, 2 routers added after the upgrade are not producing the same gaps in graphs. Any ideas?

Also, my memory usage graphs for win2k3 servers using nsclient are no longer graphing at all after the upgrade.

In the log file:
WARNING: Result from CMD not valid. Partial Result: Memory usage: total:

Does the poll expect to only get numbers and therefore will choke on a response like:

Memory usage: total:5729.37 Mb - used: 704.92 Mb (12%) - free: 5024.45 Mb (88%)
total:5729.37 used:704.92

Which worked fine in version 0.8.5. I have output fields defined as 'used' and 'total'. As I said, it was working fine previous to the upgrade.

Cacti running on FC2, php 4.4.3, apache 2.0, mysql 4.0.21

EDIT: I've just added the prepatched files from http://www.cacti.net/downloads/patches/ ... e-patched/. I'll let everyone know if that fixes it.

EDIT2: Doesn't seem to have fixed the gaps from occuring.

_JPL_ · Post by **_JPL_** » Tue Nov 30, 2004 1:52 pm

I fixed the memory graph problem. It was as I thought. Too much output confused this version of cacti.

Post by **TheWitness** » Tue Nov 30, 2004 8:20 pm

__JPL__,

Based upon your findings, you can disregard my other post. Yes, Cacti 0.8.6x is very persnikety in parsing output. It wants the data to be purely numeric in the VariableA:ValueA VariableB:ValueB. This was done intentionally.

TheWitness

_JPL_ · Post by **_JPL_** » Wed Dec 01, 2004 3:08 pm

Any idea about the gaps? Are there any problems with caching that you might be aware of? I'm seeing a lot of the following errors in my log file:

12/01/2004 01:56:09 PM - CMDPHP: Poller[0] ASSERT: '90991220<90991220' failed. Recaching host '10.13.100.115', data query #8.

The data query number varies as do the hosts.

In the images below you can see what I mean by 'Gaps'. Also note that I did the upgrade on Thursday and that's when it started. (2nd image)
The next thing I need to do is match up the gap times with any suspicious errors from the same time.

castor · Post by **castor** » Wed Dec 01, 2004 4:18 pm

i've got exactly the same problem. Using 0.8.6.b with upgrade (pre 8.6.c) under fedora core 2. Only strange thing i found is :

POLLER: Poller[0] Maximum runtime of 296 seconds exceeded. Exiting

from times to times in my cacti log. everytime i launched ot by hand it last about 2 seconds, but it seems that when cron is working it "sometimes" hangs. For info, i've got similar gaps in all my graph for my computer, but not exactly the same. You can have a look at http://bired.valholl.org/cacti/ loggin as guest password castor
just wondering if someone with anything else than fedoracore2 got the same trouble....

castor · Post by **castor** » Wed Dec 01, 2004 4:27 pm

i was also wandering if a script (like the ping test?) could hang the poller, and if there a debug mode for the requests...

_JPL_ · Post by **_JPL_** » Wed Dec 01, 2004 6:20 pm

You can set logging to debug level in cacti. I've done that but haven't poured over it yet. I was hoping someone already knows the problem and the fix.

Post by **TheWitness** » Wed Dec 01, 2004 7:22 pm

Yes, poorly designed scripts can do nasty things to the system.

TheWitness

Post by **TheWitness** » Wed Dec 01, 2004 7:35 pm

Castor,

Logged onto your site. I have some questions. I am seeing a problem with your system in general. I need to get some information from you on your system. Unfortunately, I am in the eastern us. It looks like the 0.8.6c will need some more work. I am not certain what is happening with the poller though.

Questions:
cmd.php or cactid?
What version of Apache?
What version of PHP?
What version of RRD?
What version of MySQL?
What version of net/ucd SNMP?

Thanks in Advance.

TheWitness

castor · Post by **castor** » Thu Dec 02, 2004 2:35 am

atm i'm using cmd.php but i already tried vith cactid (the rpm version anyway), it does the same..

httpd-2.0.51-2.9
php-4.3.8-2.1
rrdtool-1.0.49-1.1.fc2.fr
mysql-3.23.58-9
php-mysql-4.3.8-2.1
net-snmp-5.1.1-2

i must also say hat i'm not 100% sure that i'm not using ucd-snmp, i can't find it by rpm -qa, but my snmpd.cond contained the word ucd the first time i looked at it. I'll try to find more about that today and let you know. is it possible that saying to cacti that i'm using netsnmp and using ucd-snmp make those gaps in my graph ?

Post by **TheWitness** » Thu Dec 02, 2004 6:37 am

Cactor,

Please send me a dump of your database and also, please run the poller in DEBUG mode one time and send me the log file.

E-mail address is larryjadams@comcast.net

TheWitness

_JPL_ · Post by **_JPL_** » Thu Dec 02, 2004 11:39 am

The gaps in my graphs seem to correspond to the

MDPHP: Poller[0] ASSERT: '90991220<90991220' failed. Recaching host '10.13.100.115', data query #8.

errors in my log file. Does anybody have any idea about this? Why are there so many 'recaching hosts' entries? What causes those?

SatelliteFriendly · Post by **SatelliteFriendly** » Thu Dec 02, 2004 12:10 pm

We are also experiencing gaps in our graphs on a system running core1, 8.6b, php-4.3.9 & RRDtool-1.0.49. I have debug turned full blast and have noticed one significant difference between the polling cycles that work properly and those that produce gaps. The cycles that work ok produce the following two log entry, failed cycles do not:

12/02/2004 12:55:01 PM - POLLER: Poller[0] DEBUG: About to Spawn a Remote Process [CMD: /usr/bin/php, ARGS:
-q /var/www/html/cacti/cmd.php 0 0]
12/02/2004 12:55:01 PM - POLLER: Poller[0] DEBUG: About to Spawn a Remote Process [CMD: /usr/bin/php, ARGS:
-q /var/www/html/cacti/cmd.php 2 2]

I'm not sure why this process is run twice as only one parameter being polled. Then again, there is alot I don't know about this system yet (newbie). Any idea's on this update process is not being run some of the time?

_JPL_ · Post by **_JPL_** » Thu Dec 02, 2004 1:48 pm

I've just switched to cactid v0.8.6c to see if it makes any difference at all.

EDIT: It did not make the difference I was looking for. The graphs stopped graphing completely. SO i ran poller.php manually just to jumpstart it and got this error eventually:

12/02/2004 03:24:28 PM - POLLER: Poller[0] Maximum runtime of 296 seconds exceeded. Exiting.

I found where I need to change that number in poller.php up to 1000. Perhaps this was the problem all along? Nope, it wasn't. The next time I ran it, it was back to normal. Probably a better idea to leave that setting at 300 so you know when it's time to tweak other settings.

castor · Post by **castor** » Thu Dec 02, 2004 4:07 pm

i searched a bit but did not found how to run cacti in debug mode, only place where i found that is for graph. WHat's more i have no clue on how do dumb you the database, can you give me a link where i can find the explanations ?
one more detail as i'm thinking about it is that is runnig a smp (dual processor) kernel...

Gaps in Graphs since upgrading to v0.8.6b on FC2

Gaps in Graphs since upgrading to v0.8.6b on FC2

Who is online