[LU-538] Stack overflow on i686 client at mount time

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.1.0
• Fix Version/s: Lustre 2.1.0
• Type: Bug
• Priority: Major
• Reporter: Oleg Drokin
• Assignee: Zhenyu Xu
• Resolution: Duplicate
• Severity: 3

Description

I see a lot of i686 tests failing talely in "provision-1" phase.
Looking at some of them I see pretty clear stack overflows signs with a lot of stuff going in MGC too.

Here is one example of many: https://maloo.whamcloud.com/test_logs/4d5ab19a-b786-11e0-8bdf-52540025f9af (scroll towards the end).

Additionally, Chris is working on separating mount failures into a different code and leaving provision=-1 for jut node setup before lustre gets involved at all.

Comments

[nasf (Inactive)]

It maybe duplicate of LU-539

[Oleg Drokin]

no, I think it's not.
The crash you point to is an assertion failure in MGS thread.
This one is stack overflow on client.

[nasf (Inactive)]

We are not sure it is just the duplicate of LU-539 (sorry), so reopen it, and to see whether LU-539 patch can fix this issue or not.

[Peter Jones]

Bobi

Could you please look into this one?

Thanks

Peter

[Andreas Dilger]

The stack overflows are a duplicate of LU-411, for which I just inspected a patch from Yang Sheng.

[Zhenyu Xu]

dup of LU-411