try using a number for allocation but drop the number by 1 mb per member (if using a meta). For example, if you have a 4 member meta with space of cap/mp of 34,856 when you bind to the pool, bind this as 34,852 mb.
I faced a similar issue, I was unable to create new tdev's with preallocate=all when there is clearly enough space. Several of the tasks that I tried using preallocate=all were a bit flakey. However, I am able to create without preallocated and perform a start allocate after the create.
There was a known issue regarding this which was also documented in primus emc257440. The fix states that the issue i sfixed in SE7.2.1, unfortunately we still faced the problem in 7.3.1.
The support recommended to upgrade to v7.3.2. This issue is also addressed in the 7.3.2 release notes.
As my request was urgent to created devs, i created them without binding to pool. I was able to successfully create them. Havin said that i was able to successfully create/bind/preallocate together 15GB devices. Only 10 x 240GB i was not able to bind/preallocate. I created the 10 x 240GB devs without bind using SMC.
Then I was able to bind and fully preallocate all of them , no error.
I used SMC for above.
Having said that above all worked fine from same server to different array, i was able to create/bind/fully preallocate all 15GB and 240GB devices. Strange...
Kavan2
50 Posts
1310
0
Posted February 29th, 2012 04:00
No, it was number and after checking few threds i changed it to ALL . Same issue.