I am working with support on this. They are trying to re-produce this in the lab but what they leaning towards is that it's a reporting issues in symcli. I did more tests over the weekend:
source device is 200G, all Flash VMAX
1) create snapshot
2) "link -copy" - took 5 minutes to complete according to "list -linked -detail"
3) "relink -copy" (snapshot has not changed) - took 1 minute to complete according to "list -linked -detail"
So from this it looks like it's an incremental "copy". That raises questions:
1) Why did it take 1 minute to determine that nothing changed ?
2) So now there is absolutely no reliable way to determine how much "incremental" change to be copied to target ?
3) Counters displayed during "list -linked -detail" are bogus, they are not presenting true rate, invalid tracks owned to target.
OK, thanks, well both of these fixes are on the box I used for testing. I can see that they are linked to an SR that is similar to your scenario, so I think your process should work as designed with these fixes.
good, they will be installed on my VMAX3's very soon.
Do you know how VMAX3 calculates what's needed to be copied to target device during relink -copy command ? I know it does not rely on delta between two snapshots.
There is still a delta as such to determine which tracks have changed and keeps track of the pointers to to the data in the pools for the source and snapshot devices. It is not the same method as in the past. Previously protection session bitmaps were used, but this method is superseded to allow for the increase the the possible number of snapshots.
TimeFinder/Clone is busted on VMAX3 as well. It does the same type of "scan" like snapvx, where it scans for hours (in my case of 30TB database) before anything happens. AND support has not desire of fixing it. How in the world did this make it out of the door ?
dynamox
9 Legend
•
20.4K Posts
0
February 22nd, 2016 06:00
Jasonc,
I am working with support on this. They are trying to re-produce this in the lab but what they leaning towards is that it's a reporting issues in symcli. I did more tests over the weekend:
source device is 200G, all Flash VMAX
1) create snapshot
2) "link -copy" - took 5 minutes to complete according to "list -linked -detail"
3) "relink -copy" (snapshot has not changed) - took 1 minute to complete according to "list -linked -detail"
So from this it looks like it's an incremental "copy". That raises questions:
1) Why did it take 1 minute to determine that nothing changed ?
2) So now there is absolutely no reliable way to determine how much "incremental" change to be copied to target ?
3) Counters displayed during "list -linked -detail" are bogus, they are not presenting true rate, invalid tracks owned to target.
Thoughts ?
dynamox
9 Legend
•
20.4K Posts
0
February 25th, 2016 13:00
Jasonc,
Take a look at fixes 83953 and 84247.
PedalHarder
3 Apprentice
•
465 Posts
0
February 25th, 2016 14:00
OK, thanks, well both of these fixes are on the box I used for testing. I can see that they are linked to an SR that is similar to your scenario, so I think your process should work as designed with these fixes.
dynamox
9 Legend
•
20.4K Posts
0
February 25th, 2016 14:00
good, they will be installed on my VMAX3's very soon.
Do you know how VMAX3 calculates what's needed to be copied to target device during relink -copy command ? I know it does not rely on delta between two snapshots.
Thank you
PedalHarder
3 Apprentice
•
465 Posts
0
February 25th, 2016 14:00
There is still a delta as such to determine which tracks have changed and keeps track of the pointers to to the data in the pools for the source and snapshot devices. It is not the same method as in the past. Previously protection session bitmaps were used, but this method is superseded to allow for the increase the the possible number of snapshots.
dynamox
9 Legend
•
20.4K Posts
0
April 22nd, 2016 14:00
TimeFinder/Clone is busted on VMAX3 as well. It does the same type of "scan" like snapvx, where it scans for hours (in my case of 30TB database) before anything happens. AND support has not desire of fixing it. How in the world did this make it out of the door ?