re:Fastcgi "OS error: interrupted system call" still happens

Mike Baptiste (baptiste@nortel.ca)
24 Jul 1997 12:09 EDT

Message-Id: <199707241648.MAA00362@relay.openmarket.com>
Date: 24 Jul 1997 12:09 EDT
To: fastcgi-developers@OpenMarket.com
From: "Mike Baptiste" <baptiste@nortel.ca>
Subject: re:Fastcgi "OS error: interrupted system call" still happens

We also saw the same error on our servers (Apache 1.1.3, Perl 5.002, HPUX 
9, FastCGI compiled into perl and mod_fastcgi v1.4.3).  I had to write a 
script that monitored the error logs and kicked the server automagically 
when it happened.  However, I have now moved FastCGI to a new server 
runnign Apache 1.2.1, Perl 5.004_01, and HPUX 9, FCGI-.30 w/ sfio, 
mod_fastcgi v.1.4.3)  I have yet to see the problem.

The one thing different (besides using sfio) is I applied the memory leak 
patch posted on the listserv earlier (see below)  

Don't know if this was the prob (too soon to tell) but I do know my 
processes on the old server were growing each day and now they have 
stabilized, but its too soon to tell.

Mike Baptiste
NORTEL, Inc
baptiste@nortel.ca

FcgiCleanUp doesn't free the  TString structures (DStringFree frees
the contents, but not the structure)

so after the
	DStringFree(infoPtr->header)
	DStringFree(infoPtr->errorOut)

you need to add
        Free( infoPtr->header );
        Free( infoPtr->errorOut );



In message "Fastcgi "OS error: interrupted system call" still happens", 
rikhtver@OpenMarket.com writes:

>
>Message-Id: <19970723151719.13361@clari.net>
>Date: Wed, 23 Jul 1997 15:17:19 -0700
>From: Brad Templeton <brad@clari.net>
>To: fastcgi-developers@openmarket.com
>Subject: Fastcgi "OS error: interrupted system call" still happens
>Mime-Version: 1.0
>Content-Type: text/plain; charset=us-ascii
>X-Mailer: Mutt 0.61.1
>Organization: ClariNet Communications Corp.  http://www.clari.net/brad
>
>I'm using Fastcgi to support the page bn.newspage.com, but about once
>a day we will see this error start on the server and it doesn't go
>away untill you kick it.   I have a daemon watch the logs and kick the
>web server if it sees the error -- sigh.
>
>The cgi was compiled with version 1.5.1 of the library of Dec 12 96,
>and it has the loop around the accept call described in earlier messages
>on this list as a fix to this problem, but it still shows up.
>
>The "OS error: connection refused" error also was showing up until I
>bumped the listen-queue-depth up to a high number.
>
>Running apache 1.2.0 but not the new beta mod_fcgi which was not reliable
>enough for this.
>
>It's hard to test for as it seems to be something that only shows up
>under load, though there have been times when it slowed up
>immediately.
>
>Any thoughts?  
>
>------- End of Forwarded Message
>
>
>
>