Unsolved
This post is more than 5 years old
87 Posts
0
1174
November 17th, 2015 02:00
FAST-VP, no tiering within
Hi,
we're having a VNX5300 running 05.32.000.5.209 with a couple of pools where each pool has two tiers.
One of the tiers consists of multiple SAS raidgroups in a (4+1)R5 setup but looks like there is no "within" rebalancing taking place.
Two of the RG's are hot, the one we've added a couple of weeks ago is doing next to nothing.
What might be the reason there is no rebalancing "within" taking place? FAST-VP is moving between the two tiers but not within.
The pool is 74% full.
Regards,
John
No Events found!
ZaphodB
195 Posts
0
November 17th, 2015 06:00
It should be doing what you expect; perhaps it is just not going as fast as you think.
Things to check:
On the pool properties panel, look at the tiering tab; is auto-tiering scheduled?
Check the relocation schedule, to insure that you are allowing full use of non-prime hours.
Look at the 'move within' column in the tiering details; are there GBs waiting to move?
On tiering status, what is the estimated time for data relocation?
If the process is back-logged you can consider expanding your hours for relocation temporarily to give it more time for this activity. I believe 'move up' activities take precedence over 'move within'; if it isn't finishing the moves up, it won't be getting to the moves within.
Rainer_EMC
6 Operator
•
8.6K Posts
0
November 17th, 2015 08:00
Using current VNX OE would be a good start – yours is about two years old
kelleg
6 Operator
•
4.5K Posts
0
November 18th, 2015 14:00
Re-balancing - moving slices within a single tier - only works when the slice temperature is higher on the slices in one of the internal raid groups than another internal raid group. This was added in Flare 32 for the VNX1 (VNX2 always had this). Sometime it look it should be moving slices within but the differences are not large enough. In the Pool Properties/Tiering you should be able to see this.
Adding more disks to the tier will also automatically trigger a re-balancing within the tier.
Under certain situations you may see re-balancing not taking place - it might be related to flare version, although I can't think of any updates that relate to this specifically. I would recommend upgrading your flare version due to age.
glen
DaGrin
87 Posts
0
November 19th, 2015 23:00
Here an example of two disks from two separate private RG's in the pool:
looks to me that an IOPS diff of around 100 during the day should be enough to do some re-balancing between these RG's.
Scheduling is enabled, it's tiering between tiers but tiering within a tier always looks to be 0.
Is there some logging somewhere to check the rebalancing it did the last couple of days?
Rainer_EMC
6 Operator
•
8.6K Posts
0
November 20th, 2015 03:00
How about navicli ?
see the section about autotiering in the manual
SP log would also show autotiering actions
brettesinclair
2 Intern
•
715 Posts
0
November 20th, 2015 03:00
I don't think the FAST VP algorithms are that simplistic.
As has been suggested, you really should schedule in an upgrade.