Moritz Willers wrote: > hm. I have come to notice that occasionally the metadevices drop into > "Need Maintenance" at boot without the mdmonitord running. I'm sure > that's more the "metasync" run during mdmonitor startup than anything > else (this is 06/06). Thus I do run mdmonitord.
Yes, we do monitor that and run metasync as needed. We also run 'metasync -r' on each reboot - just to be sure. Good luck. =wb > Fortunately the two have been separated in OpenSolaris and there is now > a metasync service independent of mdmonitord. mdmonitord will stay > disabled as well. > > So I hope you are running a metasync somewhere; otherwise I recommend > you check with a metastat that your metadevices are actually "Ok". > > Still not sure though, which of the commands in mdmonitord triggers > rpc/meta to be enabled and if that moved to system/metasync. And mind > you, in my testing it only seemed to enable the rpc/meta at the first > five reboots after staging and then it somehow stops. > > I will try running the boot up in debugging on Monday. Failing that > I'll have to go over to the dtrace list and ask how I could fire up the > pid provider on any processes linking against libmeta looking for calls > of meta_smf_enable ... > > > oh, all that fun with SMF and DTrace you can have in Solaris 10! - mo > > > On 15 Sep 2006, at 22:48, The Real Warren Belfer wrote: > >> Well, we don't do mirrored root disks (we use a different methodology) >> but we do use other mirrored disks on virtually all of our servers >> including metadb replicas. And I see: >> >> Data: disabled Mar_24 svc:/network/rpc/mdcomm:default >> Data: disabled Mar_24 svc:/network/rpc/metamed:default >> Data: disabled Mar_24 svc:/network/rpc/metamh:default >> Data: disabled Mar_24 svc:/network/rpc/meta:default >> Data: disabled Mar_24 svc:/system/mdmonitor:default >> Data: online Mar_24 svcs -l >> >> So the only thing online is metainit. >> >> And it does not change on reboot. Also: >> Data: disabled Mar_24 svc:/network/rpc/bind:default >> >> Maybe mirrored root disks changes things - in which case I am >> happy we do it differently. :-) >> >> >> HTH, >> >> =wb >
