Summary: | dummy connections should avoid SSL vhosts | ||
---|---|---|---|
Product: | Apache httpd-2 | Reporter: | Arkadi Shishlov <arkadi> |
Component: | Core | Assignee: | Apache HTTPD Bugs Mailing List <bugs> |
Status: | RESOLVED FIXED | ||
Severity: | enhancement | CC: | bjorn.wiberg, brian, f.malfatto, markus, takashi.asfbugzilla, uhlar |
Priority: | P2 | ||
Version: | 2.2.6 | ||
Target Milestone: | --- | ||
Hardware: | Other | ||
OS: | other |
Description
Arkadi Shishlov
2006-05-24 12:39:32 UTC
This is caused by the core trying to activate any accept filters that might leave a process stuck waiting for data from a client. I don't think it's really worth making real SSL connections to fix this; what would be useful though is that dummy_connection() could specifically look for a non-SSL vhost to use. But ->protocol is not getting set to "https" for such servers at the moment unless explicitly configured :( Fixing this is an RFE for the core, current behaviour is not really causing incorrect server operation, just some error_log spam. There are also access_log records. Not nice when monitoring status codes. *** Bug 41459 has been marked as a duplicate of this bug. *** *** Bug 41191 has been marked as a duplicate of this bug. *** I have a new 64 bit apache 2.2.4 server on Solaris 10 with openssl 0.9.8e. When I DO NOT have the ssl.conf file included and I "apachectl graceful" to apache, all my processes that are gracefully shutdown are shown by a dash (-). When I have the Include line for the ssl.conf file uncommented, the number of Gs never changes, like it is locked in the graceful shutdown. any suggestions I have a suspicion that this bug caused my apache to spawn 150 processes (as many as MaxClients would allow), thus repeatedly crashing the server (once per day). This bug makes about one request per second, I suspect causing apache to spawn multiple instances, although I don't know why it would spawn as many as MaxSpareServers and not as many as MaxClients right away. *** Bug 43978 has been marked as a duplicate of this bug. *** Is there a workaround? I don't like my access_log filled with several thousand erroneous lines a day I am seeing the same problem with apache 2.2.6 from ports on FreeBSD 6.2-RELEASE-p4 FreeBSD 6.2-RELEASE-p4 #0 . We recently upgraded to Apache 2.2 from Apache 2.0. With apache 2.0 we could do an apachectl graceful at will with no impact to those visiting the sites we host. With apache 2.2 apachectl graceful causes a large load spike which results in a disruption serving requests. We have had to limit our apachectl graceful's to times when there is low request volume. One of the workarounds is using Worker MPM I reinstalled apache 2.2.6 with the worker MPM on FreeBSD 6.2-RELEASE-p4 #0. The number of "GET /" 400 480 "-" "-" log entries have pretty much gone away, I am still seeing a few. We still see a large load spike when we graceful apache. I fixed this behaviour by putting "listen 443" before "listen 80". Is it possible that apache uses last provided port for connecting to clients? Best we can do for this is I think as committed to trunk in r726065. |