VMware Cloud Community
kapiljha
Contributor
Contributor

vsan datastore is not writable

Hi Gurus,

I have setup a VSAN cluster and I can see vsan datastore is mounted on all 3 ESXi host (5.5 U2).

When I create a VM on this datastore the task finish with "general error"

I logged on to esxi host and try to create a directory or a file it gives below error.

osfs: OSFS_CreateFile:169: mkdir not enabled, failing mkdir request (pid: [vsan], cid: 525004647fb9f28a-802da8dee9e2cba4, childName: abc)

This is happening after I increase MTU size for vmk2(for VSAN) to 9000, plus I upgrade the the MTU in vDS as well.

the  maximum packet I can ping is 8972

#vmkping -I vmk2 -d -s 8972 IP

beyong 8972 packets are not pingable

Any pointer would really help.

Thanks

Kapil

8 Replies
ezequielcarson
Enthusiast
Enthusiast

Have you tried going back to 1500 bytes on the MTU and see what the results?

I guess the 8972 is fine. That’s the maximum size of the payload. Then the net stack will add the head for ip

Reply
0 Kudos
kapiljha
Contributor
Contributor

Yups 1500 works fine.

Reply
0 Kudos
ezequielcarson
Enthusiast
Enthusiast

Have you tried creating a VSS with MTU 9000 and moving the VMK to the VSS instead of using VDS?

Reply
0 Kudos
CHogan
VMware Employee
VMware Employee

Maybe you need to increase the packet size on the physical switch too.

Take a look at this scenario - VSAN Part 29 - Cannot complete file creation operation | CormacHogan.com

It might be relevant to your situation.

HTH

http://cormachogan.com
Reply
0 Kudos
jhakapil
Contributor
Contributor

Thanks  a lot for the replies.

Turned out to be main error messages to be

"Unable to register file system c7ea7354-b046-c6b9-f0ac-002655dbcc31 for APD timeout notifications: Inappropriate ioctl for device"

so solution was to remove the host out of VSAN cluster that has UUID c7ea7354-b046-c6b9-f0ac-002655dbcc31, and join again.

Thats what VMWare support did, and its working.

Kapil

ezequielcarson
Enthusiast
Enthusiast

did u have to resyncronize the info when you rejoined the host?

i habe tons of those apd per day

Sent from my iPhone

Reply
0 Kudos
kapiljha
Contributor
Contributor

Nop, there was nothing in the datastore except for a VM (which was not appreaing in VM inventory)

After host rejoin VSAN cluster it was OK.

VMware engineer did not do anything else.

Kapil

Reply
0 Kudos
ezequielcarson
Enthusiast
Enthusiast

How did u rejoin the host to the cluster?

Esxcli vsan cluster join?

Reply
0 Kudos