CACTID Issues - Windows [RESOLVED]

mcutting · Post by **mcutting** » Wed Jun 13, 2007 2:58 am

mcutting wrote:
mcutting wrote:Cool - can I be one of the testers if you are going to put it back in ?

Thanks Larry.
Hi Larry - any possibility that an "i" version of the poller could be released that has sh.exe rolled in (as "g" does, which works fine) ?

Thanks

BUMP - Larry - any joy with this one ? Sorry - I know you are busy !

Regds

Post by **TheWitness** » Wed Jun 13, 2007 12:28 pm

In San Jose this week and in the process of buying a house. I have to work on Cactid this week likely though. I'll keep you posed.

Larry

mcutting · Post by **mcutting** » Thu Jun 14, 2007 3:21 am

Thanks Larry - much appreciated.

mcutting · Post by **mcutting** » Tue Jun 19, 2007 8:17 am

TheWitness wrote:In San Jose this week and in the process of buying a house. I have to work on Cactid this week likely though. I'll keep you posed.

Larry

Hi Larry,

Any news on the above please ?

Thanks

Post by **TheWitness** » Tue Jun 19, 2007 9:13 pm

MCutting,

I have decided to use the "cygwin" official popen as a default. The attached source and configure.ac script will implement it by default. You can still use nifty popen using the flag "--enable-nifty-popen". I want you to test without it of course.

Test this version and let me know what you find.

Larry

mcutting · Post by **mcutting** » Wed Jun 20, 2007 1:28 am

Hi Larry. Thanks for the files. I compiled these from home against the version of Cygwin I already have with no problems.

I have tested the poller, and it seems to be stable (so far). I noticed that sh.exe is still called in this version - the "G" version of the poller doesn't seem to do this - is there a reason why ?

I will continue to monitor the "I" version of the poller over the next few hours to see if it locks up or not...

Thanks

Post by **TheWitness** » Wed Jun 20, 2007 8:31 am

The reason that it continues to call sh.exe is that I have dropped nifty popen in favor of the standard "popen" in Cygwin which requires sh.exe. However, I am hoping there is some form of magic that will take place such that we don't see any more hung processes.

Could you also note if you are getting all the cmd.exe's as well? I have a theory, that you will need to confirm.

Larry

mcutting · Post by **mcutting** » Wed Jun 20, 2007 8:47 am

Hi Larry - thanks for the quick response.

I can confirm that the cmd.exe processes are running (with around 4 cmd.exe processes when the poller runs). The process sh.exe has only hung twice since I switched to the new version this morning.

Overall, this is MUCH more stable than the one on the CACTI website (compiled with latest version of Cygwin 1.5-24).

Post by **TheWitness** » Wed Jun 20, 2007 8:58 am

Well, hung at all is not acceptable...

Larry

mcutting · Post by **mcutting** » Wed Jun 20, 2007 9:29 am

When I say "hung", the poller does eventually timeout after 292 seconds (only done this twice in 9 hours, and still looking good). However, the sh.exe process doesn't close properly (I have to kill it) - but the cactid.exe processes seem to close fine.

To give you an idea, I have enclosed a screen shot of the poller stats. From 05:00 this morning, I switched to the "I" poller...

As you can see, I am averaging between 30 and 60 seconds. I'm a bit disappointed, as I was hoping to enjoy the low poller times that everyone else seems to get. Hopefully, from the graph, you can get an indication of how the poller is behaving. The peaks (292 seconds) is where I have to kill process "sh.exe".

The stats previous to 05:00 are from the "G" poller, which as you can see, is extremely stable.

Another thing that is quite annoying is that if I run the poller from the command line, it is on average 6 seconds faster than the scheduled task

mcutting · Post by **mcutting** » Thu Jun 21, 2007 1:43 am

Hi Larry,

The poller has run stable for 24 hours now. What was your theory you mentioned in the previous post ?

For the fun of it, I compiled a version of cactid with the nifty popen enabled - the strange thing is that sh.exe is still called...

Post by **TheWitness** » Thu Jun 21, 2007 3:26 pm

My theory is that when the poller times out for some reason, that when I signal things to end, that the timed out shell processes do now receive a term signal as they should.

I would like to get the Cacti logs from you system from around the two abends of the sh.exe process. If you could send them, that would be great.

Larry

mcutting · Post by **mcutting** » Fri Jun 22, 2007 3:26 am

Hi Larry,

Unfortunately, I canned the logs around this time (they were quite large). However, the poller has exceeded 292 seconds twice since we spoke last, and I do have the logs for these (although they are MEDIUM, and not DEBUG). Would these help at all ?

I should also point out that process sh.exe has not hung since I switched to the new version you kindly provided. The version on the website for the "I" based poller hangs more often than not - this is what made me choose the "G" poller instead.

Overall, CACTID seems to be quite stable so far.

Post by **TheWitness** » Fri Jun 22, 2007 6:03 am

Yes, please send them along with a copy of your stat's graph.

Larry

mcutting · Post by **mcutting** » Fri Jun 22, 2007 6:40 am

Larry - I just PM'd you these details.

CACTID Issues - Windows [RESOLVED]

Who is online