[DRBD-user] Resize DRBD / XFS without LVM

Stefan Priebe - allied internet ag s.priebe at allied-internet.ag
Wed Dec 2 11:47:21 CET 2009

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


I've internal metadata and have used the explanation on the website 1:1.

Stefan

cosmih schrieb:
> Ok, but how do you have done the growth ?
> 
> 
> On Wed, Dec 2, 2009 at 12:34 PM, Stefan Priebe - allied internet ag
> <s.priebe at allied-internet.ag> wrote:
>> Oh i don't know. Perhaps someone of DRBD devs. can say something about that.
>>
>> Stefan
>>
>> cosmih schrieb:
>>> Hi Stefan,
>>>
>>> On my setup the metadata is on a separate disk partition, so it is
>>> still needed the dumping of the metadata ?
>>> Also, the grow of the disk partition is like: delete the partition ->
>>> create a new one bigger with the same id? The bigger one need to use
>>> the same cylinders?
>>>
>>> Thank you,
>>>
>>> --
>>> cosmih
>>>
>>> On Wed, Dec 2, 2009 at 11:13 AM, Stefan Priebe - allied internet ag
>>> <s.priebe at allied-internet.ag> wrote:
>>>> Yes that's right. I was simply using the commands explained at this url :-)
>>>>
>>>>>>> 1) put the DRBD resources into secondary mode (on each server)
>>>> no simply drbdadm down resource on both
>>>>
>>>>>>> 2) stop the DRBD service on each server
>>>> no
>>>>
>>>>>>> 3) resize the disk partition used for the DRBD devices on each
>>>>> server (the disk partition used as meta-disk is not modified)
>>>> NO first dump the metadata
>>>> drbdadm dump-md resource > /tmp/metadata
>>>>
>>>> then you can grow the device - then reimport metadata and then drbd up.
>>>> Then Filesystem resize.
>>>>
>>>> Stefan
>>>>
>>>> cosmih schrieb:
>>>>> Hi Stefan,
>>>>>
>>>>> I mean the offline growing, because from the information available on
>>>>> the specified URL I understands that the online rezise is possible
>>>>> only when the DRBD device is on top of a LVM volume.
>>>>>
>>>>> Thank you,
>>>>>
>>>>> --
>>>>> cosmih
>>>>>
>>>>> On Wed, Dec 2, 2009 at 10:46 AM, Stefan Priebe - allied internet ag
>>>>> <s.priebe at allied-internet.ag> wrote:
>>>>>> Hi!
>>>>>>
>>>>>> Do you mean the offline growing or online growing? About the online
>>>>>> growing i don't know anything :-) I haven't tested it at all.
>>>>>>
>>>>>> Stefan
>>>>>>
>>>>>> cosmih schrieb:
>>>>>>> Hi Stefan,
>>>>>>>
>>>>>>> Are you so kind to detail a little bit more the growing process
>>>>>>> (because on the URL provided is vague) ? The steps are something likes
>>>>>>> below ?
>>>>>>> 1) put the DRBD resources into secondary mode (on each server)
>>>>>>> 2) stop the DRBD service on each server
>>>>>>> 3) resize the disk partition used for the DRBD devices on each server
>>>>>>> (the disk partition used as meta-disk is not modified)
>>>>>>> 4) start the DRBD service on each server (at this step the two DRBD
>>>>>>> devices will be into secondary mode)
>>>>>>> 5) put one DRBD device into primary mode
>>>>>>> 6) resize the filesystem on top the primary DRBD device
>>>>>>>
>>>>>>>
>>>>>>> Thank you,
>>>>>>>
>>>>>>> --
>>>>>>> cosmih
>>>>>>>
>>>>>>> On Tue, Dec 1, 2009 at 2:56 PM, Stefan Priebe - allied internet ag
>>>>>>> <s.priebe at allied-internet.ag> wrote:
>>>>>>>> Hi!
>>>>>>>>
>>>>>>>> I've done an offline migration cause - we don't use LVM and without LVM
>>>>>>>> online migration is not possible.
>>>>>>>>
>>>>>>>> So i can only tell you that the offline one
>>>>>>>> (http://www.drbd.org/users-guide/s-resizing.html) is working fine :-)
>>>>>>>>
>>>>>>>> Stefan
>>>>>>>>
>>>>>>>> cosmih schrieb:
>>>>>>>>> Hi Stefan,
>>>>>>>>>
>>>>>>>>> Do you succeed on your XFS, on top of DRBD, growth attempt ?
>>>>>>>>> I am curious on this because I also need to do a growth of my DRBD device.
>>>>>>>>> Here is my setup:
>>>>>>>>> 1) Heartbeat standby/active over DRBD secondary/primary setup
>>>>>>>>> 2) DRBD 8.0.16
>>>>>>>>> 3) HW RAID 10 --> sda6, 350GB disk partition; sda7, 1GB disk partition
>>>>>>>>> --> drbd0, DRBD device on sda6; meta-disk on sda7  --> LVM PV --> one
>>>>>>>>> LVM VG --> two LVM LV --> ext3 (meaning that the DRBD device use as
>>>>>>>>> its storage a disk partition and there is LVM on top of DRBD and there
>>>>>>>>> is ext3 on top of LVM)
>>>>>>>>> What I need ?
>>>>>>>>> Basically, I need more space on DRBD device, meaning from 350GB to
>>>>>>>>> 450GB and I have this free space on HW RAID 10 volume.
>>>>>>>>> Because the DRBD setup is used by some applications who need a very
>>>>>>>>> high uptime I am interested by a solution for this growth who use the
>>>>>>>>> secondary/primary feature of DRBD and the standby/active feature of
>>>>>>>>> heartbeat or, at least, need a very low downtime.
>>>>>>>>>
>>>>>>>>> I would appreciate any advices.
>>>>>>>>>
>>>>>>>>> Thank you,
>>>>>>>>> --
>>>>>>>>> cosmih
>>>>>>>>>
>>>>>>>>>> Hi!
>>>>>>>>>>
>>>>>>>>>> Yes of cause HW Raid.
>>>>>>>>>>
>>>>>>>>>> I'll do a test today - i've already prepared testequipment yesterday.
>>>>>>>>>>
>>>>>>>>>> Stefan
>>>>>>>>>>
>>>>>>>>>>> Stefan Seifert schrieb:
>>>>>>>>>>> On Tuesday 24 November 2009 11:03:11 you wrote:
>>>>>>>>>>>>> No you haven't. Like newer versions of fdisk you can use partprobe to
>>>>>>>>>>>>> tell the kernel to re-read partition tables.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Regards,
>>>>>>>>>>>>> Stefan
>>>>>>>>>>>> Thanks Stefan for your answer. I know partprobe - but does it work on a
>>>>>>>>>>>> mounted partition?
>>>>>>>>>>> For all I know it should work with a mounted partition as well. If the
>>>>>>>>>>> partition were not mounted or otherwise used, you wouldn't need partprobe.
>>>>>>>>>>>
>>>>>>>>>>> And I assume "mounted" in this context means "used as drbd storage device".
>>>>>>>>>>> I also assume that by RAID you mean some hardware RAID, because partitioning
>>>>>>>>>>> an MD RAID wouldn't make much sense.
>>>>>>>>>>>
>>>>>>>>>>> Like with all these things it's a very good idea to first test it on a test
>>>>>>>>>>> system. Ideally identical machines, but if not available at least some VM (we
>>>>>>>>>>> use qemu for that and though its pretty slow, its enough for such tests). Life
>>>>>>>>>>> gets so much more relaxed, if you don't have to experiment around with your
>>>>>>>>>>> production machines :)
>>>>>>>>>>>
>>>>>>>>>>> Regards,
>>>>>>>>>>> Stefan
>>>>>>>>> _______________________________________________
>>>>>>>>> drbd-user mailing list
>>>>>>>>> drbd-user at lists.linbit.com
>>>>>>>>> http://lists.linbit.com/mailman/listinfo/drbd-user



More information about the drbd-user mailing list