Re: [SLE] Strange raid init sequence - revisited

Some time ago, on 03.02.25 we were talking about a strange sequence of events when mounting a software raid:

...

On Monday 24 February 2003 15:11, Carlos E. R. wrote: [...] The sequence is this: the kernel loads, initialises a lot of things, like IDE, checks the partitions, sees some of them are raid, and tries to load

The 03.02.24 at 21:44, Paul Uiterlinden wrote: them: [but fails the first time] [ ...] Now the root filesystem has been really mounted, and the raid is again considered:

<6>Freeing unused kernel memory: 168k freed <6>md: Autodetecting RAID arrays. <6> [events: 00000344] <6> [events: 00000344] <6>md: autorun ... <6>md: considering hda11 ... <6>md: adding hda11 ... <6>md: adding hdb14 ... <6>md: created md0 <6>md: bind <6>md: bind <6>md: running: <hda11><hdb14> <6>md: hda11's event counter: 00000344 <6>md: hdb14's event counter: 00000344 <6>md: RAID level 1 does not need chunksize! Continuing anyway. <6>md: raid1 personality registered as nr 3 <6>md0: max total readahead window set to 508k <6>md0: 1 data-disks, max readahead per data-disk: 508k <6>raid1: device hda11 operational as mirror 0 <6>raid1: device hdb14 operational as mirror 1 <6>raid1: raid set md0 active with 2 out of 2 mirrors <6>md: updating md0 RAID superblock on device <6>md: hda11 [events: 00000345]<6>(write) hda11's sb offset: 3124544 <6>md: hdb14 [events: 00000345]<6>(write) hdb14's sb offset: 3124544 <6> [events: 0f227faa] <3>md: invalid raid superblock magic on md0 <4>md: md0 has invalid sb, not importing! <4>md: no nested md device found <6>md: ... autorun DONE.

Now I'm noticing that it says "invalid raid superblock magic on md0". I

Well, that's a extract of what we were saying then. I have discovered some new info. I was watching the console 10 when halting the system, and noticed a message about md being stoped, and something that failed - unfortunately, the system powers off at that precise moment, and I can not read it. I have edited "/etc/init.d/halt" right at the end. First, I added a "sleep 20": echo "----------- real halt comming in 20\" ----------" sleep 20 echo "----------- real halt now ----------------------" # Now talk to kernel exec $command -d -f I do see those lines on the output, and I read a message about md being stopped. But that's all! The error message I see goes after that "exec $command -d -f" (probably halt -p -d -f), so I still can not read it. It's probably doing a "halt -p -d -f": -f Force halt or reboot, don't call shutdown(8). -d Don't write the wtmp record. The -n flag implies -d. -p When halting the system, do a poweroff. This is the default when halt is called as poweroff. So, I remove the "-p" (not shown), and this is what I see on console number 10, copied by hand: analog.c [...] gameport md: recovery thread got woken up ... md: recovery thread finished ... ---> kernel halt is called here md: stopping al md devices md: marking sb clean md: updating md0 RAID superblock on device md: hda11 [events: 000004c2]<6>(write) hda11's sb offset: 3124544 md: hdb14 [events: 000004c2]<6>(write) hdb14's sb offset: 3124544 md: md0 switched to read_only mode flushing ide devices: hda hdb hdc hdd System halted My guess is that the software raid is trying to do something as a result of the halt command, but can not finish writing it to disk. Thus, on the next wakeup, there are always errors: <6>md: created md0 <6>md: bind <6>md: bind <6>md: running: <hda11><hdb14> <6>md: hda11's event counter: 000004c2 <6>md: hdb14's event counter: 000004c2 <6>md: RAID level 1 does not need chunksize! Continuing anyway. <6>md: raid1 personality registered as nr 3 <6>md0: max total readahead window set to 508k <6>md0: 1 data-disks, max readahead per data-disk: 508k <6>raid1: device hda11 operational as mirror 0 <6>raid1: device hdb14 operational as mirror 1 <6>raid1: raid set md0 active with 2 out of 2 mirrors <6>md: updating md0 RAID superblock on device <6>md: hda11 [events: 000004c3]<6>(write) hda11's sb offset: 3124544 <6>md: hdb14 [events: 000004c3]<6>(write) hdb14's sb offset: 3124544 <6> [events: d20feae5] <3>md: invalid raid superblock magic on md0 <4>md: md0 has invalid sb, not importing! <4>md: no nested md device found <6>md: ... autorun DONE. Interesting... I could be wrong, because you see that the event log is indeed incremented after boot (or perhaps it is so because this time I had disabled the power off), so something got written (and the offset is exactly the same every time). But a few more people have reported that "invalid raid superblock" and events messages, so something weird is still happening. -- Cheers, Carlos Robinson