4 Replies Latest reply: Aug 26, 2005 1:30 PM by 807581 RSS

    iPlanet 6.0 write_error and read_timeout errors

    807581
      I'm having a lot of trouble with read timeouts and write errors in my production environment. The environment is composed of two IPlanet web servers with one cluster of two machines using Weblogic. The details follows:

      Operating System (all machines): SunOS 5.8

      WebLogic machines: WebLogic Server 6.1 SP2
      JVM Java HotSpot Client VM build 1.3.1_01, mixed mode

      IPlanet machines: iPlanet WebServer Enterprise 6.0 SP4
      Solaris VM (build Solaris_JDK_1.2.2_10, native threads, sunwjit)

      The iPlanet machines work as a redirector of requests. All requests that reach them get redirected to the WebLogic cluster, via a plugin installed on both machines. My problem is that a lot of read timeout/write error problems are occuring in this process, in various situations. The majority happens while trying to fetch static resources, but there are cases the error occurs while trying to access web server applications.
      From the IPlanet log, I selected the following errors, as examples:

      Static resources:
      [23/Mar/2004:00:05:06] failure ( 7877): for host 200.100.227.193 trying to GET /weblogic/services/images/torpedo_bg.gif, wl-proxy reports: exception occurred for backend host '10.112.46.27/7001': 'WRITE_ERROR [os error=32,  line 1379 of proxy.cpp]: '

      [23/Mar/2004:13:24:49] failure (21248): for host 200.233.164.3 trying to GET /weblogic/services/images/bt_smsmt_enviar.gif, wl-proxy reports: exception occurred for backend host '10.112.46.27/7001': 'READ_TIMEOUT [os error=22,  line 198 of Reader.cpp]: no read after 300 seconds'

      [23/Mar/2004:13:24:49] failure (21248): for host 200.233.164.3 trying to GET /weblogic/services/images/bt_smsmt_enviar.gif, wl-proxy reports: exception occurred for backend host '10.112.46.28/7001': 'READ_TIMEOUT [os error=22,  line 198 of Reader.cpp]: no read after 300 seconds'

      [23/Mar/2004:13:24:49] failure (21248): for host 201.4.214.97 trying to GET /weblogic/services/javascript/util.js, wl-proxy reports: exception occurred for backend host '10.112.46.28/7001': 'READ_TIMEOUT [os error=22,  line 198 of Reader.cpp]: no read after 300 seconds'



      Applications:
      [23/Mar/2004:00:02:50] failure ( 7877): for host 200.151.51.25 trying to POST /weblogic/services/smsmtresult.jsp, wl-proxy reports: exception occurred for backend host '10.112.46.28/7001': 'WRITE_ERROR [os error=32,  line 1379 of proxy.cpp]: '


      [23/Mar/2004:13:24:48] failure (21248): for host 200.211.229.10 trying to GET /weblogic/services/smsmt.jsp, wl-proxy reports: exception occurred for backend host '10.112.46.27/7001': 'READ_TIMEOUT [os error=22,  line 198 of Reader.cpp]: no read after 300 seconds'

      [23/Mar/2004:13:24:50] failure (21248): for host 201.1.97.212 trying to GET /weblogic/services/servlet/imageservlet, wl-proxy reports: exception occurred for backend host '10.112.46.27/7001': 'READ_TIMEOUT [os error=22,  line 198 of Reader.cpp]: no read after 300 seconds'

      The log sometimes shows an interesting message also:
      [23/Mar/2004:00:19:53] warning ( 8202): The server configuration may require more file descriptors than the operating system provides. If you encounter PR_PROC_DESC_TABLE_FULL_ERROR errors, you may wish to increase the operating system hard file descriptor limit from 10633 to 10637 (see your operating system documentation) or decrease one or more of the following settings: MaxFiles in nsfc.conf and ConnQueueSize, RqThrottle, and MaxKeepAliveConnections in magnus.conf.

      I've also checked the Weblogic nodes log, just to see if I could find an explanation to those problems. No problems were logged at the time the IPlanet reported them.

      I don't have any idea right now that could explain what's happening. Maybe the Weblogic cluster is not providing the necessary resources in time? Or is there something I could change in the IPlanet's configuration so that the frequency of the errors diminish?

      Any ideas will be greatly appreciated.

      Thanks a lot for your time,
      Alexandre