We have a peculiar case that our Zope server just stops responding after about a week, and the server gets about 50,000 hits a day. It's not using any CPU, and it's not running out of memory. It accepts connections, but doesn't follow through by returning any results. Sorry, we don't have any netstat info - we'll capture that next time it happens. Anyone have any ideas? Anywhere we can look in medusa to figure what it's doing? Richard
Was there anything noteworthy in the Z2.log right before it stopped responding? This behavior would be consistent with that of a ZODB database connection leaking every once in a while, until seven of them are leaked after which Zope sorta sits there stubbornly with its arms crossed not doing anything at all. You could prove or disprove this by making it happen again while "big M" logging was turned on... if you analyze the big M log with requestprofiler and you see "hangs" in the output that aren't caused by a restart, these are likely thread leaks. Richard Jones wrote:
We have a peculiar case that our Zope server just stops responding after about a week, and the server gets about 50,000 hits a day.
It's not using any CPU, and it's not running out of memory. It accepts connections, but doesn't follow through by returning any results.
Sorry, we don't have any netstat info - we'll capture that next time it happens.
Anyone have any ideas? Anywhere we can look in medusa to figure what it's doing?
Richard
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
-- Chris McDonough Zope Corporation http://www.zope.org http://www.zope.com "Killing hundreds of birds with thousands of stones"
If you are using a database, you might see if there are any threads in an uncompleted transaction when the sight appears to be hung. We had all kinds of problems with the exact same symptom. The database thinks one of the connections is in the middle of a transaction and so freezes all other transactions that have table dependancies. Switching from ZPoPy to Zpsycopg cleared up the issue for us. On Thursday 06 December 2001 07:07 pm, Richard Jones wrote:
We have a peculiar case that our Zope server just stops responding after about a week, and the server gets about 50,000 hits a day.
It's not using any CPU, and it's not running out of memory. It accepts connections, but doesn't follow through by returning any results.
Sorry, we don't have any netstat info - we'll capture that next time it happens.
Anyone have any ideas? Anywhere we can look in medusa to figure what it's doing?
Richard
_______________________________________________ Zope maillist - Zope@zope.org http://lists.zope.org/mailman/listinfo/zope ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope-dev )
-- Systems Analyst/Programmer VistaShare LLC www.vistashare.com
On Saturday 08 December 2001 06:06, Chris Kratz wrote:
If you are using a database, you might see if there are any threads in an uncompleted transaction when the sight appears to be hung. We had all kinds of problems with the exact same symptom. The database thinks one of the connections is in the middle of a transaction and so freezes all other transactions that have table dependancies. Switching from ZPoPy to Zpsycopg cleared up the issue for us.
Nope, straight ZODB. Thanks for the punt though :) Richard
Richard Jones writes:
We have a peculiar case that our Zope server just stops responding after about a week, and the server gets about 50,000 hits a day.
It's not using any CPU, and it's not running out of memory. It accepts connections, but doesn't follow through by returning any results.
Sorry, we don't have any netstat info - we'll capture that next time it happens.
Anyone have any ideas? Anywhere we can look in medusa to figure what it's doing? Unfortunately, no.
Just a confirmation that we have had this problem, too. Mostly on lazy almost unloaded sites. The problem (almost) disappeared, when we regular checks availability. Solaris 2.6, Zope 2.1.6 (yes an oldy)... Dieter
participants (4)
-
Chris Kratz -
Chris McDonough -
Dieter Maurer -
Richard Jones