On 5 mar, 08:04, "Ulrich Windl" <[email protected]> wrote: > On 3 Mar 2010 at 2:53, Julián Lliteras wrote: > > > Hi > > > I have some trouble while managing iscsi volumes. I have a host > > virtual server attached to iscsi volume with open-iscsi package. The > > host server have a guest virtual server allocated in the volume > > formatted with ext3 (and is the only initiator connected to). The host > > is a debian lenny. At this stage all is running ok. > > > Next, I simulate a power down problem (rebooting the iscsi target > > only). In result, the guest virtual server hangs few seconds after. > > You do something wrong: If you power down the target system, you must > power down the initiators as well, because there will be data loss at > the target, but the initiators don't realize that. You MUST replay the > journals of ext3 to make the filesystem consistent. >
You are right, I must never power down target before guests, but I will suppose that the target had a power failure or even a kernel hang (ufff). So initiators don't have any trouble, and would?!? still running with a disk failure as far. My principal worry is about data consistency, of course, the ext3 must be repaired, but in some cases the while partition has been broken and must reformat again the whole disk, lossing all data inside. Would you say that ext3 is not suitable for iscsi environments? (of course, one initiator at a lun) I tried OCFS2 filesystem and the result is an initiator absolute hang (don't know why), but has no data corruption!, entire fs was clean after power failure. I intend to determine what is the cause of data corruption and avoid that. I need to know how data is kept in this environment (and the worse thing that can happend without disk hardware failure). And how iscsi know handle possible errors like that. > > When target become online again, the virtual server never boot again > > and the host server throw volume filesystem problems. In many > > situations the volume had to be reformatted. > > > How can avoid this catastrophic situation? I know that unmounting the > > volume before disconnecting is required to avoid data corruption, but > > wat happend in a power failure ?!??! > > Not everything that can be done is a valid test. (I remember someone > putting the date forward on a production system, just to test for year > 2000 compatibility. Then some of his jobs deleted files that were > considered "very old". When he set the time back again, little things > worked as (naively) expected). > I know that all tests are not really imating reality, but as I said before I only want to study the simptoms of a target failure (where important data would be) > Regards, > Ulrich Thanks por replying Regards Julian (Yartax). -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
