Re: too many open files/Solaris/CLOSE_WAIT
"Victor Trevino Trevi?o-Alvarado" <[email protected]>
| Newsgroups | gmane.comp.web.webobjects.admin |
|---|---|
| Message-ID | <[email protected]> |
We use Solaris and the history is similar. We think that the problem is MULTIFACTORIAL. We have been study this problem and we have related with high traffic when the machine is on high load. Also, it has to do with how old the instances are, so set-up schedule is helpful. We have changed solaris parameters rlim_fd_cur=8192 and rlim_fd_max=8192, it is essential. It is also dependent on the time of request-response cycle, therefore If your app is slow it could cause it. We have seen as well the size of the queues, timeouts and many other parameters that influence the behaviour. Unfortunately we have not determined the magical solution but we have worked-around the problem and minimized it. Other problem we have (since 10 months ago) is that some instances response ports keep in CLOSE_WAIT state indefinitely. We have did many efforts to resolve this and unfortunately we couldn't at all. I think that this should be a problem between the webserver adaptor and instance adaptor when some timeout occurr. We have revised the WOAdaptor code (using a java decompiler) and we have inshights but without solutions yet. We would appreciate some help in this issue. /\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\ mc,ing.Víctor M. Treviño A. [email protected] [email protected] /\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\ >From: David Teran >To: Simon >CC: Pierre Frisch , WebObjects Admin >Subject: Re: too many open files / WO 5.2.2 Panther 10.3.2 >Date: Thu, 29 Jan 2004 15:06:17 +0100 > >Very important is that only -1- JavaMonitor instance is running. If >two instances on different servers edit the same 'wotaskd' config >cluster then strange things can happen. >On 29.01.2004, at 14:53, Simon wrote: > >> >>we've seen this problem when running multiple apps on a panther >>server and re-boot is the only way out we have found. another >>panther-related problem we see randomly is this: >> >>Cannot write to file >>/Library/WebObjects/Configuration/SiteConfig.xml. Possible >>Permissions Problem. >> >>... when trying to shut down an instance. however, if we keep >>clicking the "shut down all instances" button then it does >>eventually go away and the instances do shut down. >> >>simon >> >> >>On 28 Jan 2004, at 21:39, Pierre Frisch wrote: >> >>>David, >>> >>>I would disagree with you on this one. I agree that on WO this can >>>be home grown however in this case there is a problem in Panther. >>>I know because I have had a machine with a bunch of application >>>running on Jaguar that was very stable. The only thing that has >>>changed is the OS and the version of WO and now I have all sorts >>>of troubles. >>> >>>There are two problems on 10.2.3 one with Apache and one with >>>wotaskd. The Apache one is fairly old: the adaptor does not close >>>the connections to the app immediately and let them linger with >>>consume some of those open file. In addition wotaskd appears to >>>have developed a similar disease on Panther. The combination of >>>both is lethal. You can reduce the wotskd problem by disabling the >>>1085 config distribution feature and use teh file config. >>> >>>Pierre >>>-- >>>Pierre Frisch >>>sPearWay Ltd. >>>sPearCat Web Catalogs, the flexibility of custom design, thhe price >>>of a package. >>>http://www.spearcat.com/ >>> >>>On Jan 28, 2004, at 12:46, David Teran wrote: >>> >>>>Hi, >>>> >>>>On 28.01.2004, at 21:39, Jürgen Rohrbach wrote: >>>> >>>>>hello, >>>>> >>>>>i have read all the postings with "too many open files" but i >>>>>have still a problem to find the correct solution/workaround. >>>>>after 2-3 days my app has the following in the log file: >>>>> >>>>>IOException: too many open files >>>>> >>>>> >>>>>which solutions work to get rid of this shit! >>>>> >>>> >>>> >>>>best is to check your code. It turns out that the 'shit' is >>>>homegrown. >>>> >>>>Things to check: >>>> >>>>1) wrong usage of Process instances >>>>2) open streams >>>>3) unhandled exceptions >>>>4) wrong deployment >>>>... >>>> >>>>This might annoying for you but the problem is in your code. Do >>>>you know the lsof command or are you working on windows? >>>> >>>>cheers David >>>> >>>>_______________________________________________ >>>>WebObjects-admin mailing list >>>>[email protected] >>>>http://www.omnigroup.com/mailman/listinfo/webobjects-admin >>>> >>>_______________________________________________ >>>WebObjects-admin mailing list >>>[email protected] >>>http://www.omnigroup.com/mailman/listinfo/webobjects-admin >>> ><< DavidTerán.vcf >> > >_______________________________________________ >WebObjects-admin mailing list >[email protected] >http://www.omnigroup.com/mailman/listinfo/webobjects-admin ---------- Add photos to your e-mail with MSN 8. Get 2 months FREE*. _______________________________________________ WebObjects-admin mailing list [email protected] http://www.omnigroup.com/mailman/listinfo/webobjects-admin