[OmniOS-discuss] Oddity following upgrade to r151022

[Andy Fiddaman]

I've upgraded some of my servers from r151020 to r151022 over the weekend.
Straightforward 'pkg update --bename', no zones on these ones.

On the first boot I saw some messages about various SVM related services
(and rex) being removed due to lack of manifest, so far so good.

SunOS Release 5.11 Version omnios-r151022-f9693432c2 64-bit
Copyright (c) 1983, 2010, Oracle and/or its affiliates. All rights reserved.
Loading smf(5) service descriptions: 5/5
Delete service svc:/network/rpc/mdcomm as there are no supporting manifests
Delete service svc:/network/rpc/meta as there are no supporting manifests
Delete service svc:/network/rpc/rex as there are no supporting manifests
Delete service svc:/network/rpc/metamed as there are no supporting manifests
Delete service svc:/network/rpc/metamh as there are no supporting manifests
Delete service svc:/system/mdmonitor as there are no supporting manifests
Delete service svc:/system/metasync as there are no supporting manifests

but svc:/system/metainit:default is still there and is maintenance on
all servers.

bonnet# svcs -x
svc:/milestone/multi-user-server:default (multi-user plus exports milestone)
 State: offline since Sun May 14 11:09:51 2017
Reason: Unknown.
   See: http://illumos.org/msg/SMF-8000-AR
   See: init(1M)
Impact: 1 dependent service is not running.  (Use -v for list.)

svc:/system/metainit:default (SVM initialization)
 State: maintenance since Sun May 14 11:09:56 2017
Reason: Start method failed repeatedly, last exited with status 127.
   See: http://illumos.org/msg/SMF-8000-KS
   See: metainit(1M)
   See: /var/svc/log/system-metainit:default.log
Impact: This service is not running.

As you might expect, the start method is failing because the method script
is no longer there (/sbin/sh[1]: exec: /lib/svc/method/svc-metainit: not found)

There are some weird manifest-related properties in the service, could
these be why the service has stuck around after the upgrade?

bonnet# svcprop metainit | grep manifest
manifestfiles/code_omnios-151006_illumos-omnios_usr_src_cmd_lvm_util_metainit_xml astring /code/omnios-151006/illumos-omnios/usr/src/cmd/lvm/util/metainit.xml
manifestfiles/lib_svc_manifest_system_metainit_xml astring /lib/svc/manifest/system/metainit.xml

although I seem to have these on a few other services too, all
with 'omnios-151006' as part of the name.

I can delete the metainit service by hand but if I do that then
filesystem-local fails, seemingly because it tries to mount ZFS
filesystems in the wrong order. When I fix that (zfs umount; zfs mount -a)
then multi-user-server stays down with reason = Unknown.
That is because of rpc/smserver being in maintenance state. Deleting that
one clears everything and reboots seem ok.

Does anyone have any idea what's going on here? The same thing has happened
on four separate servers.

Thanks,

Andy

-- 
Citrus IT Limited | +44 (0)870 199 8000 | enquiries at citrus-it.co.uk
Rock House Farm | Green Moor | Wortley | Sheffield | S35 7DQ
Registered in England and Wales | Company number 4899123