On Tue, 03 Feb 2009 11:46:35 +0100 Tomasz Chmielewski <mangoo at wpkg.org> wrote: > FUJITA Tomonori schrieb: > > (...) > > >>> 2) you use slow network in this configuration, as you said before? > >> Yes. > >> Slow connection is just 5 kB/s in both directions. > >> It's easy to reproduce the problem this way. > >> Slow link is the third target above. > > > > So the connections between the target box and 192.168.4.52 are slow > > while the connections between the target box and 192.168.111.173 is > > not slow. > > > > Right? > > Yes, it is correct. > But I'm sure the same will happen with only one target connected (using > a slow link). OK, I guess that I need to see if I can reproduce it. Is there a good way to emulate a slow link? > >>> 3) All your bug reports about this problem happened in this > >>> configuration? If not, please tell me about the other configuration. > >> In "real life" it happens under slightly different (more complex, harder to set up) conditions: > >> - tgtd uses DRBD resource as the block device (backing store) > >> - link is about 100 kB/s (writes; DRBD replication over a slow internet connection) and fast local HDD IO (reads) > >> - when DRBD exchanges bitmaps with the remote peer, it suspends IO for several seconds > > > > The problem might be related with TMF. > > > > > >> - more targets configured, more initiators connected > > > > Hmm, just with lots of initiators? Even with decent link speed and no > > suspended I/Os? > > _And_ with DRBD and slow link. > With lots of initiators and decent speed link, no suspended IOs, it > behaves stable. Ok, so all the cases are with 'DRBD and slow link. > > I'm not sure about what 'more targets configured' mean. > > I mean that "tgtadm --op show --mode target" will output 40-50 targets. > 4-5 initiators connected to each target. > But I think it's not relevant, as these segfaults don't relate to the > number of initiators or targets. Yes, it's not related. Please start a new thread about this. Probably, it's easier to fix this. -- To unsubscribe from this list: send the line "unsubscribe stgt" in the body of a message to majordomo at vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html |