Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hello, The machine just use drbd device for some services but not all. Also it uses the disks dedicated to it - if that disk causes IO error for example, other disk might be still OK for all other services. There are many uses scenarios of drbd which is not good to say: if drbd device got error, shutdown or reboot . I think the better approach would be to disable drbd on it - the service/app using it will get IO error but that is it. (it might cause machine to be in high IO wait load, etc.., but this would be picked up by nagios and admin will check and then act upon it- that is better and more controllable way). I know the event handler is meant to suit everybody's need but the default behavior should not do like that - to save some unexpected surprises. cheers On Wed, Nov 30, 2011 at 8:04 PM, Felix Frank <ff at mpexnet.de> wrote: > Hi, > > On 11/30/2011 12:09 AM, Steve Kieu wrote: > > Even if IO error you may not want to halt the whole machine - best thing > > probably is to disabled/remove the drbd resource in question rather than > > kill the machine. > > that's assuming that the kernel can just "remove" a device in any given > situation, which normally (from my experience) isn't the case. > > Furthermore, if your Primary is the Sync Target, i.e. its local disk is > inconsistent, and it loses the connection to the UpToDate data, then > there is no sane way that DRBD can treat an incoming write request. > > That's why you typically want a reboot, as nothing good can come from an > inconsistent standalone primary. > > HTH, > Felix > -- Steve Kieu -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20111201/36885dce/attachment.htm>