This post is more than 5 years old
4 Posts
0
3334
Possible to avoid a full rebuild for distributed devices after a planned storage outage?
We have a VPLEX Metro (GeoSynchrony 5.5.2.02.00.03) with VNX5200 arrays as back-end storage. We will be performing data-in-place conversions on the storage arrays one at a time (VNX5200 Block to VNX5400 Block). One conversion will take about 6-8 hours if everything goes OK and during this time that storage array will be offline.
Is there any way to avoid a full rebuild for the distributed devices in a case like this?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
July 13th, 2018 08:00
Hello Salmih ,
Kindly to be noted that to avoid full rebuild , you have to maintain the logging volumes status to be OK and available , you don't want to lose them .
The EMC best practice is to keep the logging volumes mirrored , each mirror on different array for redundancy in case you lost one of the arrays.
sathyaprakash.mohandass
41 Posts
1
August 30th, 2017 08:00
Hi ,
The rebuild of distributed devices is based on the logging volume. When one leg of the volume is failed or wan link down... the volume is marked for logging. So, it is usually a logging rebuild for the distributed devices and not a full rebuild. But this also depends on the logging volume availability and I/O's that will be tracked on the logging volume and the logging volume getting full. If the logging fails, it will mark the volume as out of date and will cause a full rebuild.
In this case, I believe if there are not too many I/O's and the logging volume being online during the outage time frame and during the rebuild time, it should do a logging rebuild and not a full rebuild. I don't see any other alternative to force a logging rebuild. Unless its a new distributed device where if you don't specify the source leg it does not do a rebuild at all considering the data at both sides to be the same.
Please let us know how it goes.
salmih
4 Posts
0
June 21st, 2018 00:00
We performed the VNX conversions successfully and quicker than we had anticipated. Each array was offline for only a little over two hours during its conversion. VPLEXes on both sites were on all the time. All servers were cross-connected so no downtime was needed.
And yes, VPLEX did a logging rebuild after both conversions. Took about two hours to sync all volumes. Only the VPLEX metavolumes did a full rebuild.
ankur.patel
286 Posts
0
July 12th, 2018 10:00
Did you mirror your logging volumes across both vplex arrays using a backend cross connect?