[DRBD-user] Re: Resync of data after Primary crash

Lars Ellenberg lars.ellenberg at linbit.com
Wed Nov 14 11:33:12 CET 2007

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


> Hello,
> 
> I've read the publications about drbd 8 and 0.7 and couldn't understand one 
> thing. If anyone can clarify this to me I will be most grateful.
> 
> So here is the deal:
> 
> I am using drbd 0.7.21 (debian etch) and the test scenario is as follows...
> 
> I have a device that is 512MB in size and have configured the al-extents 
> parameter to be 15 (60MB of active space).
> 
> Now, I do a sequential write to the entire file system on the primary node. I 
> then unmount the filesystem. Then I crash the primary node (remove power) and 
> immediately power it back on (without failling over to the secondary - the 
> secondary does no writes at all).
> 
> When the machine boots it says it has to resync the entire active area of 60 
> megabytes.
> 
> The thing I don't understand is: aren't extents that have successfully 
> completed the IO operation removed from the active set?

no.

> Are these removed ONLY when new extents have to become active?

yes.

> Is this to minimize disk writes to the metadata device?

yes.

> During the primary crash there is NO active IO to the device, but the
> active set seems to remember the last extents that have been changed
> and still remembers them after all IO is complete if there was no IO 
> afterwards.

yes.

don't like it?
send a patch :)

> Please CC any reply to me, as I am not subscribed to the list...

and that is why you are not allowed to post, actually...

-- 
: Lars Ellenberg                            Tel +43-1-8178292-55 :
: LINBIT Information Technologies GmbH      Fax +43-1-8178292-82 :
: Vivenotgasse 48, A-1120 Vienna/Europe    http://www.linbit.com :



More information about the drbd-user mailing list