Bug 1205840
Summary: | Capsule/Pulp: many "Resetting dropped connection" messages on sync | ||
---|---|---|---|
Product: | Red Hat Satellite | Reporter: | Corey Welton <cwelton> |
Component: | Installation | Assignee: | satellite6-bugs <satellite6-bugs> |
Status: | CLOSED WONTFIX | QA Contact: | Katello QA List <katello-qa-list> |
Severity: | high | Docs Contact: | |
Priority: | unspecified | ||
Version: | Unspecified | CC: | akrzos, andrew.schofield, bkearney, cpaquin, cwelton, cyril.cordouibarzi, egolov, ehelms, fcami, ktordeur, mhrivnak, mjahangi, mmccune, parmstro, sauchter, stbenjam, tcarlin |
Target Milestone: | Unspecified | Keywords: | ReleaseNotes, Triaged |
Target Release: | Unused | ||
Hardware: | Unspecified | ||
OS: | Unspecified | ||
Whiteboard: | |||
Fixed In Version: | Doc Type: | Bug Fix | |
Doc Text: | Story Points: | --- | |
Clone Of: | Environment: | ||
Last Closed: | 2018-09-04 18:01:49 UTC | Type: | Bug |
Regression: | --- | Mount Type: | --- |
Documentation: | --- | CRM: | |
Verified Versions: | Category: | --- | |
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
Cloudforms Team: | --- | Target Upstream Version: | |
Embargoed: | |||
Bug Depends On: | |||
Bug Blocks: | 1190823 |
Description
Corey Welton
2015-03-25 18:12:50 UTC
Since this issue was entered in Red Hat Bugzilla, the release flag has been set to ? to ensure that it is properly evaluated for this release. Found in version Satellite-6.1.0-RHEL-7-20150320.1 It appears that setting MaxRequestsPerChild to 0 in: /etc/httpd/conf.d/prefork.conf MaxRequestsPerChild 0 this problem appears to go away. The theory is we are maxing out the # of connections to the Satellite when we execute a node sync, setting this value to 0 is essentially unlimited. To clarify, the MaxRequestsPerChild is set to 0 on the *Satellite*, not the Capsule It is unclear to me if there is truly a casualty here other than possibly higher cpu usage. If KeepAlive is turned off you will see this message rapidly. (Every download is a new connection which runs more cpu usage than reusing connections with KeepAlive.) Perhaps the build tested in this BZ did not have KeepAlive turned on yet. With KeepAlive On you will see less of this message depending on the capacity of your Apache server. Your milage will vary dependent upon hardware and the number of capsules/clients making requests against the Satellite 6.1 server. Pulp by default is tuned to 5 threads downloading content while synchronizing, thus the more capsules/clients you have downloading content, the greater potential in saturating the maximum number of processes Apache will allow by configuration to serve content. (The default config has ServerLimit/MaxClients set to 256, everything else is queued on ListenBacklog) There are several Apache Tunables which come into play here: /etc/httpd/conf/httpd.conf: KeepAlive MaxKeepAliveRequests KeepAliveTimeout /etc/httpd/conf.d/prefork.conf: StartServers 8 MinSpareServers 5 MaxSpareServers 20 ServerLimit 256 MaxClients 256 MaxRequestsPerChild 4000 # MaxConnectionsPerChild (Apache 2.4 - Accepts both) Is there a specific issue we are seeing with syncing in addition to the log line dumped to /var/log/messages? >Is there a specific issue we are seeing with syncing in addition to the log line dumped to /var/log/messages?
I don't know, other than it *appears* that it does this for a long time before sync commences. But I don't know for sure.
Also see this with Satellite 6.2 GA in QCI case#01687634 - having the same error message while syncing with satellite 6.2 server Error: Aug 23 16:48:31 capsule server pulp: requests.packages.urllib3.connectionpool:INFO: Resetting dropped connection: satellite server I know that message looks bad, but I think it's actually normal and can safely be ignored. It is not an error. A polite http client will maintain a connection to a web server and make multiple requests over the same connection. Depending on the server's configuration, it may not want to allow a connection to stay open indefinitely. Servers are commonly configured to close a connection in one or more of these circumstances: - the connection has been idle for some period of time - the connection has already been used to handle some configurable maximum number of requests - the process handling the connection has already handled some configurable maximum number of requests - the server process is being gracefully restarted When the connection gets closed, that's no problem. The client expects this to happen and just makes a new connection for the next request. I think all of the cases cited here of having seen the "dropped connection" message can be explained by this very normal behavior. The message we are seeing is just a python library noting that a connection was dropped, and thus its connection pool is responding accordingly. I definitely see how the wording of the message appears problematic, but it can be ignored. Michael, whose client reports this connections? I could move this to be an rfe in that project. The log statements come from the python "urllib3" library. https://github.com/shazow/urllib3 Since the discussion has indicated a possible change to Satellite httpd configuration, I am going to move this bugzilla over to Installer to investigate. Created redmine issue http://projects.theforeman.org/issues/16954 from this bug Thank you for your interest in Satellite 6. We have evaluated this request, and we do not expect this to be implemented in the product in the foreseeable future. We are therefore closing this out as WONTFIX. If you have any concerns about this, please feel free to contact Rich Jerrido or Bryan Kearney. Thank you. The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days |