Zum Inhalt springen
BlogLinodeLassie, der Shutdown-Wachhund

Lassie, der Abschalt-Wachhund

[b]Lassie - Linode Autonomous System Shutdown Intelligent rEbooter[/b]

Lassie ist ein Shutdown-Watchdog. Er startet Ihre Linode automatisch neu, wenn sie sich unerwartet abschaltet. Er funktioniert, indem er erkennt, dass sich Ihr Linode ausgeschaltet hat, ohne dass ein Shutdown-Job für den Ausschaltzustand verantwortlich war. Um eine Schleife zu verhindern, gibt er auf, wenn innerhalb von 15 Minuten 5 Neustarts stattgefunden haben.

Sie ist standardmäßig aktiviert, aber die Einstellung kann auf der [url=https://www.linode.com/members/utilities/watchdog.cfm]Lassie-Seite[/url] geändert werden, die sich auf der neuen Unterregisterkarte [url=https://www.linode.com/members/utilities/]Dienstprogramme[/url] befindet.

Ein Nebeneffekt davon ist, dass nun "reboot" von innerhalb Ihrer Linode wie erwartet funktioniert, aber ein "shutdown" von innerhalb Ihrer Linode führt dazu, dass Lassie anspringt und einen Boot ausgibt. Zuvor führten beide Befehle zu einem Shutdown. Ich kann nicht gewinnen :). Ein Workaround für dieses Problem ist es, einen Shutdown-Auftrag über Lish auszuführen, indem man einen Alias für "shutdown" anlegt: "ssh LinodeUsername@hostXX.linode.com shutdown".

Benachrichtigungen per E-Mail sind geplant.

Viel Spaß!
-Chris


Kommentare (10)

  1. Author Photo

    [quote:f17891f4fb=”caker”]
    A side effect of this is that now "reboot" from inside your Linode will function as expected, but a "shutdown" from inside your Linode will result in Lassie kicking in and issuing a boot. Previously both of those commands resulted in a shutdown. I can’t win :). [/quote]

    For what it’s worth, I think this a better default. I might reboot from inside the linode to make sure new libraries are in use, or just invoke the nostalgia of my WIndows day, but the only reason I shutdown is to re-arrange disks or somesuch, which I’m going to be doing from the Web UI anyway.

  2. Author Photo

    [quote:df0090ef37=”caker”][b]Lassie – Linode Autonomous System Shutdown Intelligent rEbooter[/b]

    Lassie is a shutdown watchdog.
    [/quote]

    Since it’s [i]almost[/i] Halloween, please rename it to:

    [b]L[/b]inode [b]A[/b]utonomous [b]S[/b]ystem [b]S[/b]hutdown [b]I[/b]ntelligent r[b]E[/b]booter ag[b]E[/b]nt [b]A[/b]nd [b]T[/b]im[b]ER[/b] (LASSIE EATER)

    James

  3. Author Photo

    I’m a little curious — have shutdowns been a problem?

    I’ve been on Linode since ’04, and mine has never gone down on its own.

    I’m not trying to diss a new and valuable feature — but is it solving a problem that people actually have?

  4. Author Photo

    [quote:c80aeb5696=”astrashe3″]I’m a little curious — have shutdowns been a problem?[/quote]
    Definitely not widespread. But, Linux has bugs from time to time, and so does UML. It’s just a nice thing to have, rather than having your Linode go down 5 minutes after you go to bed, and not knowing about it until you get up the next morning. This also solves the "It’s been 2 years since I’ve logged into linode.com, lost my password, the email address you have for me no longer works, and I typed reboot in my Linode and it didn’t come back up" issues (because I hack UML to shutdown on "reboot" so it’ll pick up any config changes between boots).

    -Chris

  5. Author Photo

    I got a pretty major problem when I went to resize my main drive with the new disk space we’ve just been given. Lassie started booting before it finished and pretty nastily barfed the drive.

    Lassie should check that another job isn’t already running before forcing a boot!

  6. Author Photo

    Lassie doesn’t boot your Linode — it just queues up a boot job. Jobs run one after another. And, the resize jobs *should* still be checking that your volume isn’t in use before doing anything… I’ll have a look at the logs.

    -Chris

  7. Author Photo

    OK, I found, and squashed a bug. It was a race. We use the pid file that UML writes out in determining if a Linode is running or not. It turns out that UML doesn’t immediately write this file. If another job ran just after the Linode booted and then checked to see if the Linode was active, it would not find the pid file and assume it was ok to do its thing.

    Now boot jobs wait a few seconds, while checking to see if UML wrote out its pid file, before completing.

    This bug has been in here for a LONG time (way before Lassie), so I’m grateful to you for reporting this, but sorry it ended up in your filesystem being munched on a bit.

    Regards,
    -Chris

  8. Author Photo

    [quote:2a883d7a6d=”caker”]
    This bug has been in here for a LONG time (way before Lassie), so I’m grateful to you for reporting this, but sorry it ended up in your filesystem being munched on a bit.
    [/quote]

    That’s cool. It seems to have e2fscked without any real problems. My only issue now is that the Linode dashboard reports the fs as 20 gigs but inside the linode it’s still 16 gigs. I suspect it might need some manual command-line hackery to get it back in shape, but not sure where to start.

    Feel free to bring my linode down if you think you can resize it to its proper size. Otherwise, any ideas on how to do it?

  9. Author Photo

    Try shrinking it to 16G (or 15G) and then re-enlarging? Also, what does fdisk show for the partition size? It’s possible that the partition got enlarged, but not the file
    system. If that’s the case, then you should be able to enlarge the file system from within the linode.

  10. Author Photo

    [quote:2a7ae073e2=”SteveG”]Try shrinking it to 16G (or 15G) and then re-enlarging?[/quote]

    That did the trick. Thanks!

Kommentar abgeben

Ihre E-Mail Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit *gekennzeichnet