Page 1 of 2

LSFS Updated

Posted: Wed Dec 03, 2014 4:50 pm
by Alex (staff)
Dear Beta-team members, we have reworked defragmentation process in LSFS devices:
- Defragmentation algorithms changed to handle data growth better.
- Defragmentation can be started manually with context menu command of LSFS device.

Initial implementation of defragmentation algorithm had too high threshold value for starting defragmentation process. Defragmentation process had low priority. In a result, defragmentation could not keep pace with data writes on device in some cases. This led to uncontrollable growth of LSFS journal files on physical disk.
We have raised threshold value that starts fragmentation to 40% of journal file. When more than 60% of data file is overwritten with new journal entries, system starts defragmentation process for the file.
Priority of defragmentation process has been lifted up too. When disk space is low, defragmentation process throttles down client writes to execute clean up more efficiently.

Beta build is available for download from the following link:
https://tmplink.starwind.com/StarWindBeta.exe

Re: LSFS Updated

Posted: Thu Dec 04, 2014 5:52 pm
by rrbnc
This is great! Any idea when this will be released as a new update for production?

Re: LSFS Updated

Posted: Fri Dec 05, 2014 9:51 am
by Anatoly (staff)
1-2 weeks should be enough if new issues will not be found.

Re: LSFS Updated

Posted: Fri Dec 05, 2014 2:59 pm
by KLSsandman
Hi,

Thats good news, does it fix any other issues with LSFS, performance/mount time?

Many thanks
Simon

Re: LSFS Updated

Posted: Fri Dec 05, 2014 4:07 pm
by anton (staff)
What's wrong with performance? Do you have any numbers to share?
KLSsandman wrote:Hi,

Thats good news, does it fix any other issues with LSFS, performance/mount time?

Many thanks
Simon

Re: LSFS Updated

Posted: Mon Dec 08, 2014 8:18 pm
by wgeddes
Ah great! Setting up a test SQL environment now for testing. Thanks

Re: LSFS Updated

Posted: Wed Dec 10, 2014 2:10 pm
by Anatoly (staff)
Thanks for the heads up!

Re: LSFS Updated

Posted: Tue Dec 16, 2014 9:30 am
by KLSsandman
Does the latest 7471 build have the fixes included in the beta as the beta version number is older than this?

Thanks
Simon

Re: LSFS Updated

Posted: Tue Dec 16, 2014 9:41 am
by Anatoly (staff)
No, it is not, unfortunately. By the way, you can always check the release notes here:
http://www.starwindsoftware.com/release-notes-build

Re: LSFS Updated

Posted: Thu Dec 18, 2014 3:48 am
by rrbnc
Can we get emailed when new builds are released?

Re: LSFS Updated

Posted: Thu Dec 25, 2014 7:43 pm
by Anatoly (staff)
We will do even better - we have planned to change our software to search for updates automatically.

Re: LSFS Updated

Posted: Mon Jan 05, 2015 3:52 pm
by Vladimir Usov
sorry, is there any fresh update when this fix appears into production release?

Re: LSFS Updated

Posted: Mon Jan 05, 2015 5:36 pm
by glecit
I've had one issue with beta build 7450 in my 2-node HA lab cluster. I have a total of two LSFS devices. Both were working for a few days through reboots and tests. One device has had no problems, but I'll describe what has happened with the device hosting a CSV volume:

- After a clean reboot, nodeA began a sync, but got stuck at 29% and never moved beyond that. I left it over the weekend to be sure it wouldn't progress.
- I tried to restart the Starwind service, but it would not stop cleanly.
- I tried rebooting nodeA, but afterwards it displayed "Current Node is not synchronized" and never began synchronization. If I clicked on Synchronization manually, it would bring up an error that said "Failed: some partner node is already starting synchronization for current node." Meanwhile, nodeB showed "Connection Status Active. Synchronization Status Not synchronized" for the nodeA partner.
- I tried restarting nodeB, which caused synchronization to begin again, but it was stuck at the same point (29%).
- I then removed the target and device from nodeA, deleted the device files, and restarted the Starwind service. I then tried to add nodeA as a replica server for the device. The wizard returned that the device was created, but synchronization stuck at 5%. I tried the same thing again and it stuck at 21%. The next time device creation failed. The fourth time I tried, the sync completed successfully and the device is working properly again.

I've noticed a couple of things:

- Connectivity status always shows good from both ends.
- While stuck syncing, a file (CSV1_HA.swdsk.tmp) is continually created and deleted in the storage path for the device on nodeA.
- One of the times it was stuck during synchronization, the server log on nodeB was being spammed with these identical lines every 15ms:
15 93524.807 760 HA CBlocksBarrierEnterBarrier Intersects exist
15 93524.807 760 HA CBlocksBarrierEnterBarrier CurrentBlockID=0x000000000575B7F0, ullStartSector = 0x21EE8, ulLength = 45, bIsSerialize = 1
15 93524.807 760 HA CBlocksBarrierEnterBarrier OtherBlockID=0x0000000005709C20, ullStartSector = 0x21EE8, ulLength = 45, bIsSerialize = 1

I'd be happy to send logs, but didn't want to post them here.

Is this a known issue?

Re: LSFS Updated

Posted: Tue Jan 13, 2015 5:49 pm
by glecit
It's stuck during a fast sync again after a clean reboot. I'm going to leave it like this for now in case you want to do a remote session while it's stuck.

It would be nice to hear if this is a known issue or not.

Re: LSFS Updated

Posted: Sat Jan 17, 2015 11:27 am
by anton (staff)
1) Looks like snapshot based sync (well, VERY close...) so good news we've re-worked this code path in an upcoming post-fix.

2) We'd be happy to take a look over remote session as I'm still not 100% sure about 1) so please drop a message to support with a link to this thread and credentials for you RDP or whatever (TeamViewer?) account.

Thank you for cooperation! :)
glecit wrote:It's stuck during a fast sync again after a clean reboot. I'm going to leave it like this for now in case you want to do a remote session while it's stuck.

It would be nice to hear if this is a known issue or not.