8.0, journaling and UFS snapshots - better, but not perfect yet

I tried to set up journaling under 7.x and had problems dumping the journaled filesystem, since creating the snapshot would often hang or panic the machine. I decided to try again with 8.0, and while the hangs and panics have gone away, taking the snapshot still takes upwards of 5 minutes, during which time the system is all but hung. Additionally, I frequently see these:


For now, I will keep the journaling in place, since I think it does more good than harm, but I wonder if the act of creating a snapshot couldn't either be sped up, or perhaps not lock the system quite as pervasively as it does.

Note that the swap in this case is on a dedicated partition on the disk, not in a swap file on the journaled FS or anything like that (that would make those messages make sense, IMHO).

Yesterday the system paniced with a dup_alloc. I would share the dump information, but the system hung while dumping. When I restarted it, it didn't stay up for more than 30 seconds before it did it again - and again, got stuck dumping.

This time, I booted single user and forced an fsck on the journaled filesystem. It found an unlinked inode and cleared it. A subsequent reboot restored the system to normal operation.

I realize that this isn't likely to be enough information to be helpful, but unfortunately it is all the information I have because the panic dumps didn't finish normally.

did you use journaling with gjournal?

Best way to solve this is to use ZFS. Making snapshot takes about a second (seems constant; it doesn't depend on a filesystem size or number of files in it). Use "zfs send" in place of dump. Works reliably and fast.