Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Tue, Feb 26, 2008 at 11:34:21AM +0100, Andreas Haase wrote: > Hi list, > > Am Mittwoch, den 13.02.2008, 15:04 +0100 schrieb Wolfram Schlich: > > I am using drbd-8.0.10 on linux-2.6.24 with heartbeat-2.1.3. > > Sometimes, on failover, the kernel spits out one or more of > > these messages: > > > > kernel: drbd0: local disk flush failed with status -5 > > > > What does that mean? Is it harmful? > > same problem occurs here with kernel 2.6.24.2 and drbd 8.0.11. Because > there has been no answer on the list, here again the question: does that > log error message have any implications on data consistency of the drbd > volume? the log message not. the aparently not (suficiently) implemented flush may lead to data/meta data not on stable storage when we think it is. so if you have volatile write cache enabled, and you have a hard crash, there may be situations where data loss is possible, because there may be some sectors missing from the next resync. note that drbd versions before 8.0.11 did not pay attention to this at all, and just assumed that the run on either write-through, or battery backed non-volatile write-back cache. > The message is being repeated VERY often yes. a driver that does not support flush should return EOPNOTSUPP, which would cause us to log this once and no longer try it. returning EIO gets logged by drbd, each time, and we will try it the next time again. the more io, the more log entries. pick a working combination, or, after making sure it is really harmless at your site (it could still be a real error code!), filter it out in syslog-ng. > on one of our server clusters. if the other clusters are fine, what is the difference, what makes this cluster behave "bad"? -- : Lars Ellenberg http://www.linbit.com : : DRBD/HA support and consulting sales at linbit.com : : LINBIT Information Technologies GmbH Tel +43-1-8178292-0 : : Vivenotgasse 48, A-1120 Vienna/Europe Fax +43-1-8178292-82 : __ please use the "List-Reply" function of your email client.