VERY Slow Expansions

  • Last Post 28 June 2021
Jay Dubb posted this 25 June 2021

We've used Promise arrays for the past 15 years, and this is only a problem on our R2600i units.  Our 1840i units, no problem.

Example:  Container (20 TB) made up of 12 disks RAID 6, and has a 2 TB logical drive configured as an iSCSI target.  Click the option to expand that LD from 2 TB to 3 TB, commit and confirm.  Just a small expansion, tons of room available in the container for it.

On the 1840i, the expansion is very fast and the entire expansion is complete within a couple of HOURS.

On the R2600i (same 12 disk 20 TB RAID 6 config)-- a newer and faster chassis-- that exact same operation takes 2 WEEKS.

What has happened with the R2600 that makes simple operations take WEEKS when they only took HOURS on much older units?  Is there a workaround or settings change that can be made to fix this?  This is killing us...

Order By: Standard | Latest | Votes
Gautham Sakthi posted this 25 June 2021

Hello Jay,

Kindly register a web support ticket by logging into so we can check further on this issue.



R P posted this 28 June 2021

Hi Jay,

Online capacity expansion can be a complicated operation. If you have one array and one LUN with space to expand into, the resize operation will be done almost instantly, then it will re-sync.

The complication comes when you have several LUNs in the array and/or when you add disks to the array.

If you add disks to the array then the all LUNs in that array will have to be restriped across the new disks. This can take time.

If you have 3 LUNs and want to expand the first LUN (LD 0), then the other 2 LUNs will have to be moved to make room to expand the first LUN into as LUN expansion requires a contiguous free space. This also can take a lot of time.

Also, LUN expansion is a background task and it yields to user IO. The more the storage is used the longer the background task will require. This is true for any background task, including syncronization.

Jay Dubb posted this 28 June 2021

Thank you for that info.  But this brings me back to the same question.... why?  What changed?

Both the 1840i and R2600 have 12 physical disks for data-- the remaining 4 are configured as global spares.  Both have the same number of LUNs, same number of LDs, and each LD is the same size.  

Each server gets a 2 TB iSCSI volume on each of the 2 Promise chassis.  An E: on the 1840i, and an F: on the R2600.

If a server needs a 1 TB expansion on E:, then we also expand F: an identical amount, so both drives are always the same size.  (we use electronic document storage software that writes 2 copies of every record, one copy to E: and a duplicate copy to F:)

We start an expansion on both Promise units at the same time.  In 100% of the cases, 1840i is finished in a few hours and the new space is available.  R2600 takes weeks before the expanion completes and the new space is visible.

So obviously, something significant has changed in how the array expansions are handled.  But I cannot wrap my head around why such a simple amount of storage expansion takes 2 orders of magnitude longer on the *newer* hardware.  (3 hours vs. 300+ hours)