Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Thu, Oct 27, 2011 at 08:43:30AM +0200, Andrew McGill wrote: > Hi DRBD users, > > On DRBD 8.4.0 running on CentOS 6.0 64 bit, I have just run into a > non-oops-non-crash-infloop problem of some sort. Below is a stack of > possibly relevant information. Sadly, this is the second DRBD project > that has failed for me - the first was unsatisfactorily plagued by > kernel panics on real genuine RHEL around 3 years back. > > I hope there's sufficient information here to help you find the bug, So you get "hang on high IO load". And that *obviously* is the fault of DRBD, a bug that has to be fixed in drbd. Because *obviously* that's something *no one* *ever* has observerd without drbd in the stack. That's why you do not get a single hit on google for hang on high IO load. Right. Not. I'm not saying that DRBD is bug free. It certainly is not. But even if you can only reproduce something with DRBD in the stack, that may just because DRBD causes additional stress on memory, busses, and IRQs. The traces you provide suggest that DRBD is waiting for completion of IO (meta data transactions, in this case) to the local disk, which for some reason does not happen. It may also be some DRBD internal deadlock, or something that DRBD would be supposed to at least work around. Hard to tell from the available information. You could also try to *reproduce* it, and try to reproduce it with the latest DRBD 8.3.x as well as the 8.4.0. Just, please do not jump to conclusions. > while I approach my project in a completely different manner. But yes, sure, that's your call. Cheers, -- : Lars Ellenberg : LINBIT | Your Way to High Availability : DRBD/HA support and consulting http://www.linbit.com