[ovirt-users] Problem with data domain link after update to 3.6.2

[Joop]

On 27-1-2016 10:55, Gianluca Cecchi wrote:
> Hello,
> single physical host with oVirt 3.6.1 on CentOS 7.2
> Configured with SH Engine on NFS.
> Also NFS iso and NFS data domain present
> All 3 NFS are provided by host itself.
> 2 VMs with CentOS 7.2.
>
> Starting point has all ok in webadmin GUI, also the hosted_storage
> without exceptions.
>
> Updated to 3.6.2 following these steps:
> On host
> - global maintenance
>
> On engine VM
> - yum update "ovirt-engine-setup*"
> - engine-setup
>
> On host
> - exit maintenance
>
> Verified correct access to webadmin GUI again. From GUI
> - shutdown of the 2 powered on VMs
>
> On host
> - local maintenance
>
> Shutdown of engine VM
>
> On host
> yum update
> BTW: should I stop vdsmd before running update or not?
> reboot
> exit maintenance
>
> Verified access to webadmin GUI and all storage domains are shown as up.
> But when I try to power on one of the VMs I get this error:
>
> VM racclient1 is down with error. Exit message: Cannot access storage
> file
> '/rhev/data-center/00000001-0001-0001-0001-0000000000ec/556abaa8-0fcc-4042-963b-f27db5e03837/images/d2f2b967-c0c3-4648-9daa-553e8ef4652b/e0524893-1f78-4e6b-8e5e-83217bc0dc5b'
> (as uid:107, gid:107): No such file or directory.
>
> Indeed the file, that actually seems to be a symbolic link, doesn't exist
> On host:
> [root at ractor ~]# ll /rhev/data-center/00000001-0001-0001-0001-0000000000ec
> total 0
> lrwxrwxrwx. 1 vdsm kvm 99 Jan 27 07:31
> 2025c2ea-6205-4bc1-b29d-745b47f8f806 ->
> /rhev/data-center/mnt/ractor.datacenter.mydomain.dom:_SHE__DOMAIN/2025c2ea-6205-4bc1-b29d-745b47f8f806
> lrwxrwxrwx. 1 vdsm kvm 99 Jan 27 07:31
> fd5754f1-bd00-4337-ad64-1abde35438ae ->
> /rhev/data-center/mnt/ractor.datacenter.mydomain.dom:_ISO__DOMAIN/fd5754f1-bd00-4337-ad64-1abde35438ae
> lrwxrwxrwx. 1 vdsm kvm 99 Jan 27 07:31 mastersd ->
> /rhev/data-center/mnt/ractor.datacenter.mydomain.dom:_SHE__DOMAIN/2025c2ea-6205-4bc1-b29d-745b47f8f806
>
>
> [root at ractor ~]# ll /rhev/data-center/mnt/
> total 0
> drwxr-xr-x. 3 vdsm kvm 74 Nov 19 15:47
> ractor.datacenter.mydomain.dom:_ISO__DOMAIN
> drwxr-xr-x. 3 vdsm kvm 74 Nov 19 15:46
> ractor.datacenter.mydomain.dom:_NFS__DOMAIN
> drwxr-xr-x. 3 vdsm kvm 74 Nov 19 15:25
> ractor.datacenter.mydomain.dom:_SHE__DOMAIN
> drwxr-xr-x. 2 vdsm kvm  6 Nov 19 15:25
> _var_lib_ovirt-hosted-engine-setup_tmp8InOft
>
> [root at ractor ~]# ll
> /rhev/data-center/mnt/ractor.datacenter.mydomain.dom\:_NFS__DOMAIN/556abaa8-0fcc-4042-963b-f27db5e03837/
> total 4
> drwxr-xr-x.  2 vdsm kvm   69 Nov 19 15:46 dom_md
> drwxr-xr-x. 10 vdsm kvm 4096 Dec  4 00:23 images
> drwxr-xr-x.  4 vdsm kvm   28 Nov 19 15:47 master
>
> [root at ractor ~]# ll
> /rhev/data-center/mnt/ractor.datacenter.mydomain.dom\:_NFS__DOMAIN/556abaa8-0fcc-4042-963b-f27db5e03837/images/
> total 32
> drwxr-xr-x. 2 vdsm kvm 4096 Dec 10 15:45
> 3b641c29-5196-4b2f-b1a5-fb31f8064780
> drwxr-xr-x. 2 vdsm kvm 4096 Dec  1 12:06
> 7d5dd44f-f5d1-4984-9e76-2b2f5e42a915
> drwxr-xr-x. 2 vdsm kvm 4096 Nov 25 16:34
> 83799ab5-055f-445c-8b6c-496d19bce921
> drwxr-xr-x. 2 vdsm kvm 4096 Nov 29 17:44
> 9a3fcdf7-75f4-4605-ba34-665dae9a4e0d
> drwxr-xr-x. 2 vdsm kvm 4096 Dec  4 00:19
> ba11e44c-7cf5-4ef1-a9cd-0a6630bda801
> drwxr-xr-x. 2 vdsm kvm 4096 Dec  4 00:22
> d2f2b967-c0c3-4648-9daa-553e8ef4652b
> drwxr-xr-x. 2 vdsm kvm 4096 Dec 10 15:45
> d49f6096-96a2-4ef2-926a-a54a8246d303
> drwxr-xr-x. 2 vdsm kvm 4096 Dec  4 00:24
> d9d2cd3f-0b1c-4e04-bd6a-1ffafc723344
>
> [root at ractor ~]# ll
> /rhev/data-center/mnt/ractor.datacenter.mydomain.dom\:_NFS__DOMAIN/556abaa8-0fcc-4042-963b-f27db5e03837/images/d2f2b967-c0c3-4648-9daa-553e8ef4652b/
> total 3338720
> -rw-rw----. 1 vdsm kvm 8589934592 Jan 27 06:57
> e0524893-1f78-4e6b-8e5e-83217bc0dc5b
> -rw-rw----. 1 vdsm kvm    1048576 Dec  4 00:21
> e0524893-1f78-4e6b-8e5e-83217bc0dc5b.lease
> -rw-r--r--. 1 vdsm kvm        259 Dec  4 00:22
> e0524893-1f78-4e6b-8e5e-83217bc0dc5b.meta
>
> Is this problem already known? Should it be sufficent to only create
> the symbolic link? What process/service is in charge to create the link?
>
>
I have that problem too once in a while. Restarting vdsmd will solve
your problem BUT its because vdsmd is active before your NFS is up. A
fix is to require nfs in the vdsmd service unit.

Regards,

Joop