Wandering Thoughts archives

A (Solaris 8) automounter irritation

Here's an automounter irritation that keeps biting us: on fileservers, the Solaris 8 automounter is not smart enough to notice when an underlying /export/<whatever> mount point has been unmounted.

If it gets a request for the filesystem, it 'mounts' things anyways, which then fouls up various aspects of, say, remounting the real filesystems. It is remarkably difficult to persuade the automounter to let go so that you can restore things to normality, and you can't just shoot the entire automounter without decimating the overall system.

(We think that the usual way the automounter gets a request for the filesystem is for mail delivery to need to look at someone's .forward, but the system is reasonably opaque so we're not entirely sure.)

It wouldn't be too difficult for the automounter to notice this situation and do the right thing. Even if it didn't, we'd be happy if there was some way to tell the automounter to temporarily stop serving a mount point (short of editing the automounter maps, which has various drawbacks).

And failing all that, the automounter really needs a way to be told to immediately time out and unmount some mount it's done. Maybe umount theoretically does this, but it sure doesn't seem to work reliably for us in practice. (Possibly this is due to whatever keeps getting the automounter to mount the filesystems.)

Solaris 8 DiskSuite's lack of good monitoring

One of the things that irritates me about DiskSuite in Solaris 8 is how half-hearted the provisions for automated monitoring and reporting are. I don't necessarily ask that DiskSuite have all of this pre-built, but I would like commands that had output that at least made it relatively easy to build this ourselves, and DiskSuite falls flat on this.

Metastat will tell you all of the necessary information, but it does it in a verbose, human-readable format that is sufficiently complicated and underdocumented that I get very nervous 'parsing' it with grep, awk, and other ad-hoc solutions. (When you have a sufficiently complicated format it is too easy for ad-hoc solutions to not handle some case you didn't know about and blow things up as a result.)

I'm not sure what to make of 'metastat -p'. It doesn't seem quite designed for use by other programs, since it has just enough odd features to make that challenging (like the line continuations it likes to throw in). At the same time, it shows that the DiskSuite people were at least vaguely aware of the whole issue and made some sort of half-hearted stab at dealing with it, making it more annoying that they didn't try to do more.

We have various ad-hoc solutions, of course. I suspect that every site making real use of DiskSuite has built similar ones (and that each has some bugs and oversights, because that is the nature of ad-hoc tools).

(Some casual Googling suggests that Sun's official solution may involve SNMP. Please pardon me while I snort cider.)

I find this irritating partly because Solaris is supposed to be the 'enterprise' operating system, and good monitoring is one of the things that enterprise people traditionally care about. Solaris 8 is lacking enough things as it is, so it would be nice to get something for all of this 'enterprise' minimalism.

Possibly this is improved in Solaris 9 and Solaris 10. Unfortunately we seem unlikely to move to Solaris 10 any time soon (although it would be nice).