On Wed, 6 Sep 2017, Paul Cuzner wrote: > Hi, > > One of the users of cephmetrics has seen issues where the response to > perf dump requests against an osd seems to hang. (issue #91 - > https://github.com/ceph/cephmetrics/issues/91) > > When I looked at the admin_command function in ceph[1], it looks to me > that the socket interaction is defaulting to blocking mode? So if > there is a delay in responding, the caller will see this hanging > behaviour that was described in the issue. > > If blocking mode is being used, can anyone explain why this is > preferred over a non-blocking approach with 'select'? The queries are generally very short running so there is little value in making them asynchronous. In this case, it sounds like the OSD was hung for some other reason, though, so that a thread was holding a lock longer than it should, causing the command to hang. Do you know what command was hung, or whether other commands were issued before it? sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
