Unsolved
This post is more than 5 years old
54 Posts
0
1643
Rebalancing Variance when adding new TDevs
Wondering what you guys do/recommend when adding Tdevs to a thin pool as far as 'rebalancing variance' is concerned. Value is from 1-50% I'm thinking somewhere around 10-15% would be a good value. Wondering what you all use?
5449Pr4QlS12567
46 Posts
0
July 2nd, 2011 15:00
I am not sure if you can choose the varience yourself.
Just fire
symconfigure -sid XXX -cmd "start balancing on pool= type=thin;" commit –nop
Even after this, you might see some varience. Read Primus Id emc242014
You might wan't to avoid Rebalancing in production hours.
sauravrohilla
859 Posts
0
July 10th, 2011 05:00
and 10% variance that I have written about is actually the variance caused by the microcode when writing to the Pool.This variation is caused by the ability of the microcode to skip extent allocation on the next data dev in the pool. If that device's DA is busy that DA can pass on the allocation request to the DA that owns the next data dev in the pool so extent allocation does not have a large effect on host I/O and this causes variance.
and this 10% variance can be balanced by rebalancing process.
regards,
Saurabh Rohilla
sauravrohilla
859 Posts
0
July 10th, 2011 05:00
HI Deepak,
I am the author of this primus . Rebalance variance can be changed from 1 to 50% using
symconfigure command
set pool PoolName,,]]>;
type =
[, rebalance_variance =
[, max_dev_per_rebalance_scan =
Well as far as the recommended value is concerned, we recommend whateve the default is (its 1%) but the bigger question is does it rebalance the pool to 1% variance? It depends..
regards,
Saurabh Rohilla
5449Pr4QlS12567
46 Posts
0
August 30th, 2011 05:00
Hey Saurabh
Sorry for replying so late.The E-mail Id I am registered with on ECN doesn't work now so no notification. Switched jobs
I understand what you said. So for a senario mentioned below, if am am adding 50 more Data Devs, and wan't to keep varience to default 1%. then would I need to change "Max devs per rebalance scan" from 256 to the new count of data devs i.e 472+50=522 ? What effect will changing this number make?
-bash-3.00$ symcfg -sid xxxx show -thin -pool r5_fc15k_gp4 -detail|more
Symmetrix ID: xxxxxxxxxx
Symmetrix ID : xxxxxxxxxxxxx
Pool Name : r5_fc15k_gp4
Pool Type : Thin
Dev Emulation : FBA
Dev Configuration : RAID-5(3+1)
Pool State : Enabled
# of Devices in Pool : 472
# of Enabled Devices in Pool : 472
Max. Subscription Percent : 100
Rebalance Variance : 1%
Max devs per rebalance scan : 256
Regards
DpK
sauravrohilla
859 Posts
0
August 30th, 2011 13:00
HI Deepak,
I have not seen anyone changing the max dev per rebalance scan and there are no document that talks about increasing it. So i cant be sure. May be someone else can shed some light on it.
regards,
Saurabh
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
March 4th, 2015 09:00
Hi Deepak,
The Max dev per rebalance scan (256) is the no.of devices/tdats the system would consider to rebalance in a single scan. If the default value of 256 is increased to your new tdat count of 522, it would consume more CPU to perform rebalance as it trys to rebalance more tdats per scan. Rather the default values are mostly preferred so that a couple of scans involving 256 tdats per scan would rebalance the extents. Hope this answers your question.
dynamox
1 Rookie
1 Rookie
•
20.4K Posts
0
May 14th, 2015 08:00
i have never noticed increased response times, i still prefer to do it during off business hours.
Harshak
52 Posts
0
May 14th, 2015 08:00
Is there any performance issue during pool rebalacning ? is this something we need to perform during maintenance window ?