Christine Tran wrote: > On Fri, Mar 6, 2009 at 4:53 PM, Stephen Hahn <sch at sun.com> wrote: >> * Christine Tran <christine.tran at gmail.com> [2009-03-06 21:50]: >>> On Fri, Mar 6, 2009 at 4:48 PM, Antonello Cruz <Antonello.Cruz at sun.com> >>> wrote: >>>> Do you know if the filesystem containing /lib/svc/method/svc-mdonitor is >>>> mounted when startd tries to run the method? >>> Well, it should be, shouldn't it? We asume /lib/svc/method is mounted >>> before SMF tries to run any method? Or is this assumption wrong? I >>> would guess that when SMF tries to run mdmonitor, it did not find the >>> method, from the log. >> Unless the system has separated /lib (or /lib/svc or /lib/svc/method) >> into a separate filesystem (which would be an unsupported >> configuration in general), the method should be present. Since you >> say it's a minimized system, I wonder if the package that contains >> svc-mdmonitor has been installed at all. >> > > It's installed. In fact, when we look later, the file is there, the > entire /lib/svc/method dir is there. It's not a problem to start it > later, but the real problem is on boot-up, the disks (under SDS) don't > get metasync -r run on them, so they come up in maintenance state. A > metasync fix this, and a svcadm restart, but this seems like a > problem. > > Could the disks be too fast, I wonder? These are very new T5140 with > ridiculously high RPM. But it's all on one disk, no separate FS. I don't think it is related to the speed of the disks.
If you say the method is in /lib/svc/method and that it is not a problem to start the service later, it is very possible that this service is set to run *before* filesystem/root:default It won't be the first time I see this happening. That's why I asked if the filesystem with the method was mounted at the time SDS was supposed to start. Would you mind to tell us what are the dependencies of your service, please? Thanks, Antonello
