Summary: | mod_proxy_ws_tunnel - Infinite loop on websocket connection loss | ||
---|---|---|---|
Product: | Apache httpd-2 | Reporter: | Jo <joffroy.christen> |
Component: | Other Modules | Assignee: | Apache HTTPD Bugs Mailing List <bugs> |
Status: | RESOLVED FIXED | ||
Severity: | normal | ||
Priority: | P2 | ||
Version: | 2.5-HEAD | ||
Target Milestone: | --- | ||
Hardware: | PC | ||
OS: | All |
Description
Jo
2014-02-17 11:07:46 UTC
I've detected the issue in 2.4.7, then tested and noticed the same issue in 2.4.x (2.4.8) and 2.x (2.5). (In reply to Jo from comment #1) > I've detected the issue in 2.4.7, then tested and noticed the same issue in > 2.4.x (2.4.8) and 2.x (2.5). Thanks for the report/analysis/patch. In my system I could not easily trigger the symptom, even when I killed the client it became readable and was handled by the normal read. Can you re-test this patch since your env hits it more easily? http://people.apache.org/~covener/patches/trunk-ws-disconnect.diff (In reply to Eric Covener from comment #2) > > Thanks for the report/analysis/patch. In my system I could not easily > trigger the symptom, even when I killed the client it became readable and > was handled by the normal read. > > Can you re-test this patch since your env hits it more easily? > > http://people.apache.org/~covener/patches/trunk-ws-disconnect.diff I'd to replace APLOGNO() by APLOGNO(02453) to compile it. I've successfully tested it on 2.4.8-dev and 2.5-dev. Thanks. On app killed, it logs "AH02447: err/hup on backconn" as before. On browser killed, it logs "AH02447: err/hup on client conn" as expected now. I never had "unknown event on backconn" nor "unknown event on client conn", but I just made a few tests. We're currently using Apache 2.4.9 on Win64 which seems to solve the "infinite loop on websocket connection loss" problem for us. However, what we are seeing now is entries like [Fri Mar 21 08:00:10.405224 2014] [proxy_wstunnel:notice] [pid 19932:tid 836] [client 172.29.184.113:53972] AH02447: err/hup on backconn [Fri Mar 21 08:00:42.229688 2014] [proxy_wstunnel:notice] [pid 19932:tid 788] [client 172.29.184.113:53982] AH02447: err/hup on backconn [Fri Mar 21 08:05:08.604172 2014] [proxy_wstunnel:notice] [pid 19932:tid 840] [client 172.29.184.113:54199] AH02447: err/hup on backconn [Fri Mar 21 08:05:16.780267 2014] [proxy_wstunnel:notice] [pid 19932:tid 772] [client 172.29.184.113:54210] AH02447: err/hup on backconn [Fri Mar 21 08:05:21.724793 2014] [proxy_wstunnel:notice] [pid 19932:tid 768] [client 172.29.184.113:54216] AH02447: err/hup on backconn [Fri Mar 21 08:18:26.592849 2014] [proxy_wstunnel:notice] [pid 19932:tid 788] [client 172.29.184.113:54414] AH02447: err/hup on backconn [Fri Mar 21 08:18:26.595779 2014] [proxy_wstunnel:notice] [pid 19932:tid 764] [client 172.29.184.113:54413] AH02447: err/hup on backconn [Fri Mar 21 08:26:53.173822 2014] [proxy_wstunnel:notice] [pid 19932:tid 764] [client 172.29.184.113:54463] AH02447: err/hup on backconn [Fri Mar 21 08:40:58.011922 2014] [proxy_wstunnel:notice] [pid 19932:tid 840] [client 172.29.184.113:60997] AH02447: err/hup on backconn [Fri Mar 21 08:40:58.011922 2014] [proxy_wstunnel:notice] [pid 19932:tid 772] [client 172.29.184.113:60998] AH02447: err/hup on backconn We can't find any indication so far that there would be a problem with the websocket server (Play/Netty). So it looks like this might be an issue with mod_proxy_ws_tunnel? (In reply to Martin Studer from comment #4) > We're currently using Apache 2.4.9 on Win64 which seems to solve the > "infinite loop on websocket connection loss" problem for us. However, what > we are seeing now is entries like Please open a new bug report so that it can be referenced independently. |