[linux-lvm] How do you create large numbers of LVs? (In the 1000s) Is it even possible?
Jonathan E Brassow
jbrassow at redhat.com
Thu Jul 28 22:00:01 UTC 2005
I know that a similar bug (158687) has been fixed since your version...
It is a somewhat similar bug, but I think the previously mentioned
applies to you more. Still, it couldn't hurt to update.
brassow
On Jul 28, 2005, at 4:26 PM, Ming Zhang wrote:
> I did vgchange -an vg1
>
> and then still fail
>
> [root at fc3-i386-2 ~]# lvcreate -L12M -ntv282 vg1
> VG vg1 metadata writing failed
> [root at fc3-i386-2 ~]# lvcreate -vvvvv -L12M -ntv282 vg1
> Setting global/locking_type to 1
> Setting global/locking_dir to /var/lock/lvm
> File-based locking enabled.
> Getting target version for linear
> dm version
> dm versions
> Getting target version for striped
> dm versions
> Locking /var/lock/lvm/V_vg1 WB
> Finding volume group "vg1"
> Opened /dev/sda
> /dev/sda: No label detected
> Opened /dev/md0
> /dev/md0: Failed to read label area
> Opened /dev/sda1
> /dev/sda1: No label detected
> Opened /dev/sda2
> /dev/sda2: No label detected
> Opened /dev/sda3
> /dev/sda3: No label detected
> Opened /dev/sdb
> /dev/sdb: No label detected
> Opened /dev/sdb1
> /dev/sdb1: lvm2 label detected
> /dev/sdb1: lvm2 label detected
> Read vg1 metadata (710) from /dev/sdb1 at 50688 size 65377
> Creating logical volume tv282
> Allowing allocation on /dev/sdb1 start PE 843 length 180
> Archiving volume group "vg1" metadata.
> VG vg1 metadata writing failed
> Unlocking /var/lock/lvm/V_vg1
> Closed /dev/sda
> Closed /dev/md0
> Closed /dev/sda1
> Closed /dev/sda2
> Closed /dev/sda3
> Closed /dev/sdb
> Closed /dev/sdb1
>
> i use this script to create them
>
> #!/bin/bash
> # a simple loop to create large # of LV
>
> LIMIT=300
> a=1
>
> while [ "$a" -le $LIMIT ]
> do
> lvcreate -L10M -ntv$a vg1
> free
> let "a+=1"
> done
>
> not ram problem
>
>
> [root at fc3-i386-2 ~]# free
> total used free shared buffers
> cached
> Mem: 255044 80704 174340 0 17776
> 47400
> -/+ buffers/cache: 15528 239516
> Swap: 522104 0 522104
>
> [root at fc3-i386-2 ~]# lvextend --version
> LVM version: 2.00.25 (2004-09-29)
> Library version: 1.00.19-ioctl (2004-07-03)
> Driver version: 4.4.0
>
> shall i try new version?
>
> ming
>
>
> On Thu, 2005-07-28 at 15:49 -0500, Jonathan E Brassow wrote:
>> yeah... it could be a memory issue which is causing the difference in
>> active lvs.
>>
>> It could be the fact that he starts with _inactive_ lvs that allows
>> him
>> to create so many to start.
>>
>> To see if you are experiencing the same bug, you could 'vgchange -an
>> <vol_name>' and then try to create a bunch of lvs... Then, once
>> created, try to activate them.
>>
>> brassow
>>
>> On Jul 28, 2005, at 3:38 PM, Ming Zhang wrote:
>>
>>> i think this is strange that this guy can at least create 1500 lv but
>>> fail to activate them all.
>>>
>>> here what i found is i even can not create ~300 lv.
>>>
>>>
>>> ming
>>>
>>>
>>> On Thu, 2005-07-28 at 15:32 -0500, Jonathan E Brassow wrote:
>>>> I think the problem you are seeing is similar to the one found in
>>>> bugzilla (164198). Would you be willing to add some notes there?
>>>> It
>>>> will give you a place to track the progress...
>>>>
>>>> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=164198
>>>>
>>>> brassow
>>>>
>>>> On Jul 28, 2005, at 2:18 PM, Ming Zhang wrote:
>>>>
>>>>> On Thu, 2005-07-28 at 12:09 -0700, Nathaniel Stahl wrote:
>>>>>> We'd like to be able to create a large number of LVs (potentially
>>>>>> numbering in the low thousands). I get failure after LV 226 or
>>>>>> so,
>>>>>> though - "VG VolGroup01 metadata writing failed".
>>>>>>
>>>>>> RedHat claims this should be possible with LVM2 on the following
>>>>>> web
>>>>>> page:
>>>>>>
>>>>>> http://www.redhat.com/magazine/009jul05/features/lvm2/
>>>>>>
>>>>>> I admit to being a little suprised at the 2^32 max LV claim - I
>>>>>> was
>>>>>> figuring 2^20 as the theoretical max given 2.6's 32 bit device
>>>>>> numbering
>>>>>> scheme (20 bits for minor, 12 bits for major).
>>>>>>
>>>>>> The LVM2 code, at least version 2.00.25 as distributed in FC3,
>>>>>> appears
>>>>>> to have a check that the minor number is strictly less than 256.
>>>>>> Removing this check allows for the creation of working LVs using
>>>>>> minors
>>>>>> greater than 256, but LV creation fails with the error "VG
>>>>>> VolGroup01
>>>>>> metadata writing failed" creating the 227th LV. Even with the
>>>>>> minor
>>>>>> limit in place - I can't create more than 226 LVs.
>>>>>
>>>>> i asked this question before. there is a hard coded limitation in
>>>>> lvm
>>>>> metadata, so the real number is like this, limited around 2xx. the
>>>>> limitation will be removed in near future. how near? i do not know.
>>>>> :)
>>>>>
>>>>>
>>>>>>
>>>>>> Is there a patch that allows this limit to be broken? Should I be
>>>>>> using
>>>>>> a newer version of the tools? If not currently possible, is this
>>>>>> something that will be in the near future?
>>>>>>
>>>>>> Thanks for any help/advice you can give.
>>>>>>
>>>>>> -Nate Stahl
>>>>>>
>>>>>> _______________________________________________
>>>>>> linux-lvm mailing list
>>>>>> linux-lvm at redhat.com
>>>>>> https://www.redhat.com/mailman/listinfo/linux-lvm
>>>>>> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>>>>>
>>>>> _______________________________________________
>>>>> linux-lvm mailing list
>>>>> linux-lvm at redhat.com
>>>>> https://www.redhat.com/mailman/listinfo/linux-lvm
>>>>> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>>>>>
>>>>
>>>
>>
>> _______________________________________________
>> linux-lvm mailing list
>> linux-lvm at redhat.com
>> https://www.redhat.com/mailman/listinfo/linux-lvm
>> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>
> _______________________________________________
> linux-lvm mailing list
> linux-lvm at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>
More information about the linux-lvm
mailing list