#.dmp files

1 messages · Page 1 of 1 (latest)

icy minnow
#

So I have a cluster that is ... Out of support, a gamble taken by the forces that be, and it seems our number has been called.

I have an arbitrary amount of snap vault jobs from a source cluster just vomiting at trying to either maintain relationships, or initialize new ones.
Out of 1200 relationships, 400+ failed and refused to come back after an upstream ontap update.

Last Transfer Error: Failed to start transfer for Snapshot copy "snapmirror.4199a5b9-afca-11e9-8d81-00a098daa2f9_2152117175.2026-01-21_085000". (Failed to start transfer. (Destination not supported (Operation not supported)))
                      Last Transfer Size: -

Source: 9.17.1p2
Destination: 9.14.1p2

They're within the interop matrix for snap mirror.

When viewing event log show via set diag, it triggers a rastrace dump, one of which I can't view through windbg because well. Symbols, but. Here we are.

I'm grasping at straws and legit just need to limp this stupid cluster another year.

grizzled quarry
#

Any chance your destination has DP mirrors instead of XDP?

Snapmirror show -fields type

icy minnow
#

And I say that cuz I just recreated a handful of the destinations from complete and total scratch. So all volumes are DP. All volumes are empty. All relationships are XDP

#

Even with brand new relationships, it fails with the same error

#

I don't have ARP enabled on some.
I don't have large volume support enabled on like... 99% of them
The jobs began failing generally after an up stream ontap update to 9.17.1p2.

#

But I have 433 jobs that have failed and refused to resume or initialize, but I have almost 840 plus that are still just fine

#

So I did some digging in an effort to find some root cause and dug around the .dmp files outlined in the diag level of 'event log show' for the upset jobs.
No good. Mostly just garbage because... Well... I don't have the right tools or offsets. But I've very much hit my wall here

grizzled quarry
#

Sorry man! It’s likely some new feature that was auto enabled in 9.17 that 9.14 can’t handle. Looks like you’re in the right path.
I’d focus on just one relationship that’s working and one that isn’t

Get a list (set diag; vol show -vol good|bad -instance) and place the data side by side in excel or something. See what’s different

icy minnow
#

New vols:
Brand new tested working:
Dr > lock

Not tested so far:
Prod > Dr > lock

If the above works I'm going to compare settings like you mentioned

icy minnow
#

Yeah something is fucky with my DR array.

Original:
Source => Dr => vault
No workie

Source => new Dr => new vault
Works

icy minnow
#
  • Vol move of Dr doesn't do it.
  • Bounce if final destination cluster doesn't do it
  • None of Dr vols corelate to a specific aggr or ha pair
  • Going to try and kick ic lifs next
  • Going to try and make sure my work around works before I double up 400+ vols for a few weeks
grizzled quarry
#

Just gonna throw it out there

Have you looked at all the snapmirror logs and the event logs? Usually there is something that helps indicate any issues

icy minnow
#

Had that screenshot handy, snagging the rest

#

Event log is the same entry for the last transfer error

#

When looking under diag event logs, it whines to check the .dmp log generated

#

Snapmirror.audit has nothing for this volume or relationship ID

#

Snapmirror_error is the same. Refreshing SPI to see if I got an old copy

grizzled quarry
#

What is the volume type? That message indicates the relationship is uninitialuzed

#

Is the volume type DP? Or something else?

icy minnow
#

Created new vol and relationships at destination

grizzled quarry
#

Very odd. Have you engaged support? Maybe some odd versioning issue with a snapshot

icy minnow
# grizzled quarry Very odd. Have you engaged support? Maybe some odd versioning issue with a snaps...

Heh.
I wish

This is a fas8200 that's stuck on 9.14.1P2 -- it was supposed to be replaced this year but the c-suite had sticker shock at the "hey were going to replace your snap lock cluster" pricetag.
They enjoy the safety it brings, but they do not enjoy the concept of a $2m+ pricetag

Edit: trust me, the management in the food chain who supports this are ... Annoyed they're in this spot now lol

#

So, it is very out of support except for disks