lv mirror on 2 campus

[Nguyen Anh Tien]

Hi all guru!
I do not understand this problem.
I have one lv (name lvidx) it is mirrored from 2 pv (c16t0d0 and c12t0d0). C16t0d0 is located at one campus and the other is located at another campus.
My question: Wheren one pv is broken (such as power off VA) how do my lv work? I think it try to write on both pv so that writing speed will be slower than normal status. another person say lv can detect failure pv. it will not write failure anymore if pv is broken.
Can anyone give me the offical doc about this topic. Any comment will be highly appreciate
thank.
tienna

[Ravi_8]

Hi,

If one of your pv is broke (assume c16t0d0)then replace the disk, copy the data from c12t0d0.

If any one PV breaks you get the error in system log (/var/adm/sw/syslog.log)

[Nguyen Anh Tien]

Dear Ravi
You may misunderstand. I want to ask how do my lv work? (whether it stop to write on failue pv or retry to write on this pv when this pv is fail.)

[Ravi_8]

Hi,

How you have created mirror,

if you have created mirror from c16t0d0 to c12t0d0, inthis case if c16t0d0 fails, writing on both pv stops, if c12t0d0 fails, data will be written to c16t0d0 but not to c12t0d0.

[Steven E. Protter]

Assuming you have the right connection and have properly mirrored, if one disk power fails,nothing bad should happen at all.

Then the disk comes back online there will be stale extents which will update over a period of time.

If you replace a disk that has failed then you have to redo the mirroring from scratch.

SEP

[Nguyen Anh Tien]

Thank Davi and SEP. It mean that OS will stop to write on failure pv. It will not retry to write with given times. doesn't it?

[Henk Geurts]

no. it will not retry unless you tell it the missing disk is available again.

[Nguyen Anh Tien]

hmm :=((
I can not understand myselfs why my system was slow at that time (failure time). Syslog only display 2 error message on failure pv.
Any offical doc is recommand.
tienna

[Ivajlo Yanakiev]

Yes it happen.
When you miss one disk from mirror there little time out.
After that system work with normal speed only with good disk.

Also do not reboot with one connected disk :) or you probably need one more reboot :))

[Nguyen Anh Tien]

Can anyone give me the offical doc about this topic?

[Steve White_7]

Hi,

You might want to be careful if that system is rebooted. Volume Groups require what is known as quorum to activate. You system may be configured to actviate without quorum, if not you will have to manually run vgchange in order to mount the filesystems in that volume group. The vgsync binary can be run once you get the other storage back on-line.

Cheers

[Bill Hassell]

LVM mirroring was never designed for long distance mirroring. Unless your long distance connection is 100% fibre, you will find normal operations slower than supported configurations (local disks). In simple terms, once a mirror is not functioning, LVM will mark the extents on the disabled disk as stale and will not attempt to write anything to them. This is exactly the same as a local mirror.

But that's the simple explanation. Many failures are much more complicated and highly intelligent I/O cards (such as fibre), special fibre switches (such as Brocade) and the arrays themselves can fail in strange ways, causing long delays as the driver attempts to recover from the problem. A solid failure (complete disconnect) will probably work as expected but electronic failures may cause excessive delays in the driver. The long distance between the two disks isn't helping at all.

You have to ensure that the remote disk and communication path never go down. I would not recommend running a production server in this manner due to all sorts of timing and error recovery problems. There are special disk arrays that are specifically designed for long distance redundancy.