CACTID Issues - Windows [RESOLVED]

Post by **TheWitness** » Fri Jun 22, 2007 10:00 am

Try this version. Compile it in debug first.

TheWitness

mcutting · Post by **mcutting** » Fri Jun 22, 2007 10:21 am

Done - running now.

mcutting · Post by **mcutting** » Fri Jun 22, 2007 10:24 am

Got this on the first run

06/22/2007 04:25:18 PM - PHPSVR: Poller[0] ERROR: Input Expected, Script Server Terminating

Poller then goes on to "hang" until 292 seconds is reached (nothing in the logs though, or the stackdump file)

mcutting · Post by **mcutting** » Fri Jun 22, 2007 11:08 am

Hi Larry,

I have had to go back to the "I" version of the poller I was running before - CACTID is crashing out more often than not !

I am going to clear the log, but PM'd you a copy if you'd like to review it ?

Post by **TheWitness** » Fri Jun 22, 2007 12:05 pm

Yes, this is my fault. I forget that MySQL is not thread safe. I will revise the code and get an update here quickly. Sorry about that...

TheWitness

Post by **TheWitness** » Fri Jun 22, 2007 12:12 pm

Here is the revised sql.c file.

TheWitness

mcutting · Post by **mcutting** » Fri Jun 22, 2007 12:57 pm

Hi Larry,

Not a problem - thanks for the quick reply. I'll take a look at the revised poller, and post feedback asap.

mcutting · Post by **mcutting** » Sat Jun 23, 2007 4:56 am

Hi Larry,

The "J" version of the poller is running now. I will keep a track over the weekend to see what results we get - will keep you posted.

Thanks for your help with this.

mcutting · Post by **mcutting** » Sun Jun 24, 2007 4:18 am

Hi Larry,

Thought I'd post the results of the "J" poller so far. As it stands, there hasn't been any instance where I have had to manually kill processes, so this is working well. However, the poller has exceed the 292 seconds a couple of times (see graph).

Although I compiled CACTID in debug mode, I don't seem to have a core file - any ideas ? I can post a copy of the log if required.

mcutting · Post by **mcutting** » Mon Jun 25, 2007 10:54 am

Here is an updated "J" poller status

The poller has been running error-free since Sunday morning. Enclosed is a copy of the graph for the polling process.

Post by **TheWitness** » Thu Jun 28, 2007 7:13 pm

Mcutting etal,

I have FOUND the segfault issue in cactid. My stress testing has revealed 100% success in 10k poller runs.

I will publish the revised code for testing soon enough. My goal with the next release is to support at least 10k devices and over 100k data sources.

So, when I post, that will be a preview of that single server performance. This is without the new revised architecture (that I have still not started on) that will allow you to scale Cacti to over 100k devices.

TheWitness

mcutting · Post by **mcutting** » Fri Jun 29, 2007 1:28 am

Excellent news - thanks Larry.

I'll eagerly await the new code. The debug "J" poller I am running has exhibited some overruns, but these never seem to get logged in the stackfile....

Post by **TheWitness** » Fri Jun 29, 2007 5:07 am

Mcutting,

Attached is Beta code. I still have to verify the following:

1) Concurrent script flag is working (controls concurrent scripts)
2) Spike detection is working
3) Reindex shortcutting is working

Other than that, there is a new requirement for "Thread Safe" mysql. It will not compile without it. You may have to re-configure/compile the mysql client libraries with the thread safe option. Run "./configure --help" for more information.

Regards,

Larry

mcutting · Post by **mcutting** » Fri Jun 29, 2007 5:51 am

Thanks Larry.

I am currently recompiling / configuring the MySQL libraries using

./configure --without-server --enable-thread-safe-client

Will let you know how I get on

mcutting · Post by **mcutting** » Fri Jun 29, 2007 6:49 am

Ok. MySQL reconfigured and recompiled. CACTID new source configured and compiled. Poller is now in place, and running. First pass = 24 seconds - a speed increase of 10 seconds on the predecessor !!!!

Looks good so far - I will keep an eye on things...

CACTID Issues - Windows [RESOLVED]

Who is online