Device busy when trying to restore 2

[mronquillo]

I'm trying to restore a backup from disk (we have the DBO module and the device is the adv_file type). Clones are scheduled to start immediately after a savegroup completes. A problem I've been running into is that during the cloning process (which uses the READONLY filestream to clone to tape), restores on the same device are unable to start without having to kill the clone. The client just reports that the device is busy until cloning finishes. I was under the impression that the DiskBackup Option allows for multiple read/write streams. Am I missing something here? TIA

 

[NWNinja]

Disk Backup does allow for simultaneous read/write. You can write to the R/W device at the same time as reading from the RO device. But if you are cloning and restoring at the same time, you are trying to do two reads from the RO device at the same time, and that does not work.

 

[605]

This statement is simply wrong. As you can write multiple streams to a device at the same time you can also read multiple streams from a device at the same time. The device type does not matter.

However, to be able to do this, you must run multiple recover processes and one GUI does not allow this. But if you start multiple recover processes from the command line (for example), this works perfectly. This is basic NW functionality and it worked with NW for years, even for the old 'normal' file device type.

(Automatic) cloning is a bit tricky but it should not block the media/device for other processes to access it as long as they are of the same basic functionality (either read or write). On an advanced file type device you may even read and write at the same time due to a combined write & read functionality.

If you do automatic cloning, this happens at the end of the savegroup, so after all client backups have been done. This is why a client should not do any backup any more. If you only have one group, i expect a bug. Besides that i have no idea with this very basic information.

 

[NWNinja]

Well, ok, let´s put it this way. You can clone and restore at the same time yes, but not from the same RO device. If you are cloning from the RO device, you will need to perform your restore from thh R/W device using command line recover.
Use mminfo to locate which ssid/cloneid reside on the R/W device and supply them in the command. But the fact still remains, you cannot clone and recover at the same time from the same RO device.

From Legato:
Solution ID: legato74534
Known Issue: LGTpa63781 Request For Enhancement(RFE)
This RFE will be implement in NetWorker Keystone Release.

Presently, You can have a Write Operation (Backup) with only one Read Operation (Recovery, Manual Cloning or Staging). For more details, please refer to Table 41, Concurrent Operations with Advanced File Type Device on Page 259 in NetWorker Administrator's Guide:

http://web1.legato.com/infodev/publications/NetWorker/UNIX/7.0/uxag_7.0.pdf

Here is the problem or goal:
Cannot have Cloning operation on an 'Advanced File Type' device already performing a recover session
Cannot recover from the adv_file device when a Clone operation is reading from the same adv_file device
Error: nsr server busy recover waiting 30 seconds then retrying

Causes of this problem:
You can not have a Read Operation (Cloning) with another Read Operation (Recovery)

 

[605]

Good point - NW will in fact accept more than one read process but it will not accept any clone process if a read is already in process. This will be queued until the read terminates (NW 7.2).

 

[mronquillo]

Thanks for the info, NWNinja and 605. We finally have a "workaround" with the RO file device is in use.

 

[getvell]

can you post the work around!!
RV

 

[mronquillo]

It's basically what NWNinja said.

Use mminfo to get the cloneid for the saveset you want to restore. Then use some form of "recover -S ssid/cloneid", using the RW device instead of the RO device (which is busy when cloning).

 

[getvell]

I have a problem with that!!. The problem is the fact that in an ADV_File type device we can only read from the RO device, we cannot read from the write device.

So when we have a 1TB of data to be staged off to tapes from the ADV_File type device and the staging is in process a restore call during that won't succeed until the staging process is killed.

I discussed this issue with an SE from Legato and agrees that this is a bad design from Legato and they suggested having more smaller ADV_File type devices and do some performance tuning which i don't completely agree with them.
-RV