Loading ...
Sorry, an error occurred while loading the content.

Re: [linux] Kernel 2.6.14 - is it stable?

Expand Messages
  • Godwin Stewart
    ... Hash: SHA1 On Mon, 7 Nov 2005 11:01:22 +0100, Godwin Stewart ... The machine just locked up again (using 2.6.14), but I was able to
    Message 1 of 22 , Nov 7, 2005
    • 0 Attachment
      -----BEGIN PGP SIGNED MESSAGE-----
      Hash: SHA1

      On Mon, 7 Nov 2005 11:01:22 +0100, Godwin Stewart <gstewart@...>
      wrote:

      > > I suspect it might just be a 'blip' I've known X to lock up at times.
      >
      > But it wasn't just X that locked up

      The machine just locked up again (using 2.6.14), but I was able to do the
      SysRq-s/u/b thing so no damage was done to the filesystem.

      There's nothing in the logs that might give a clue.

      I'm sticking with 2.6.13.4 for now.

      - --
      G. Stewart - gstewart@...

      God prefers spiritual fruit, not religious nuts...
      -----BEGIN PGP SIGNATURE-----
      Version: GnuPG v1.4.1 (GNU/Linux)

      iD8DBQFDb18/K5oiGLo9AcYRAkd6AJ9RIt5tygFiVuJx+1ze42bsSbhM6ACghmdU
      iDPgmOyH0Eaaf+KrgMoehSk=
      =Ad5W
      -----END PGP SIGNATURE-----
    • ed
      On Sun, 6 Nov 2005 18:37:18 -0600 ... It s part of my evil masterplan. I don t want to have a single point of failure, beit a NFS server for example. I just to
      Message 2 of 22 , Nov 7, 2005
      • 0 Attachment
        On Sun, 6 Nov 2005 18:37:18 -0600
        "Rivera, Angel R" <Angel.R.Rivera@...> wrote:

        > Why do you want a distributed file system?
        > Why raid 0/1?

        It's part of my evil masterplan. I don't want to have a single point of
        failure, beit a NFS server for example. I just to have a single volume
        on an array of boxes, in both replication and load balance. I have
        thought about rsync, but that means I have a single write node, which
        could die, then I have down time. It's this single point... and yes,
        switches can melt etc, but for the moment switches are acceptable as
        failure points.

        Perhaps another reason, 'baby wants rattle'. Windows does DFS quite
        well, but because of license costs I cannot afford to provide this with
        samba mappings.

        --
        Regards, Ed http://www.usenix.org.uk - http://irc.is-cool.net
        A TCP/IP stack was the worst feature windows ever got
        ~
        ~
        :wq
      • Rivera, Angel R
        ... From: linux@yahoogroups.com [mailto:linux@yahoogroups.com] On Behalf Of ed Sent: Monday, November 07, 2005 12:54 PM To: linux@yahoogroups.com Subject: Re:
        Message 3 of 22 , Nov 7, 2005
        • 0 Attachment
          -----Original Message-----
          From: linux@yahoogroups.com [mailto:linux@yahoogroups.com] On Behalf Of
          ed
          Sent: Monday, November 07, 2005 12:54 PM
          To: linux@yahoogroups.com
          Subject: Re: [linux] Kernel 2.6.14 - is it stable?


          On Sun, 6 Nov 2005 18:37:18 -0600
          "Rivera, Angel R" <Angel.R.Rivera@...> wrote:

          > Why do you want a distributed file system?
          > Why raid 0/1?

          It's part of my evil masterplan. I don't want to have a single point of
          failure, beit a NFS server for example. I just to have a single volume
          on an array of boxes, in both replication and load balance. I have
          thought about rsync, but that means I have a single write node, which
          could die, then I have down time. It's this single point... and yes,
          switches can melt etc, but for the moment switches are acceptable as
          failure points.

          Perhaps another reason, 'baby wants rattle'. Windows does DFS quite
          well, but because of license costs I cannot afford to provide this with
          samba mappings.

          ----

          you are aware that when a node goes down that is part of a dfs or gfs
          everything stops on that file system, aren't you?
        • ed
          On Mon, 7 Nov 2005 13:58:32 -0600 ... Don t know about Windows DFS like that, we ve had problems. gfs is something different, AFS is ok, it s got a sensible
          Message 4 of 22 , Nov 7, 2005
          • 0 Attachment
            On Mon, 7 Nov 2005 13:58:32 -0600
            "Rivera, Angel R" <Angel.R.Rivera@...> wrote:

            > you are aware that when a node goes down that is part of a dfs or gfs
            > everything stops on that file system, aren't you?

            Don't know about Windows DFS like that, we've had problems.

            gfs is something different, AFS is ok, it's got a sensible design with
            backup volumes. But it's only got one write node, and thats what I
            really don't like.

            --
            Regards, Ed http://www.usenix.org.uk - http://irc.is-cool.net
            A TCP/IP stack was the worst feature windows ever got
            ~
            ~
            :wq
          • Rivera, Angel R
            ... Don t know about Windows DFS like that, we ve had problems. gfs is something different, AFS is ok, it s got a sensible design with backup volumes. But it s
            Message 5 of 22 , Nov 7, 2005
            • 0 Attachment
              -----Original Message-----
              Don't know about Windows DFS like that, we've had problems.

              gfs is something different, AFS is ok, it's got a sensible design with
              backup volumes. But it's only got one write node, and thats what I
              really don't like.
              ---

              for your needs it is about the same. It will still break as soon as
              one of the dfs/gfs nodes goes down and hang the other nodes up. A GFS
              may survive it tho that portion of the data will be unavailable until
              it is rebuilt.

              I am using raid5 at home with pretty good success. We use it here as
              well.
            • Shawn Tan
              ... could a coda fs do the trick?? cheers.. -- with metta, Shawn Tan. It is easier to change the specification to fit the design than vice versa.
              Message 6 of 22 , Nov 7, 2005
              • 0 Attachment
                On Monday 07 November 2005 18:53, ed wrote:
                > It's part of my evil masterplan. I don't want to have a single point of
                > failure, beit a NFS server for example. I just to have a single volume
                > switches can melt etc, but for the moment switches are acceptable as
                > failure points.

                could a coda fs do the trick??

                cheers..

                --
                with metta,
                Shawn Tan.

                It is easier to change the specification to fit the design than vice versa.
              • ed
                On Mon, 7 Nov 2005 14:39:09 -0600 ... Although slow, I m eventually going to look at MySQLfs. MySQL 5 has clustering of it s own, so I don t see why not to
                Message 7 of 22 , Nov 7, 2005
                • 0 Attachment
                  On Mon, 7 Nov 2005 14:39:09 -0600
                  "Rivera, Angel R" <Angel.R.Rivera@...> wrote:

                  > for your needs it is about the same. It will still break as soon as
                  > one of the dfs/gfs nodes goes down and hang the other nodes up. A GFS
                  > may survive it tho that portion of the data will be unavailable until
                  > it is rebuilt.
                  >
                  > I am using raid5 at home with pretty good success. We use it here as
                  > well.

                  Although slow, I'm eventually going to look at MySQLfs. MySQL 5 has
                  clustering of it's own, so I don't see why not to make use of it's
                  replication and balancing. Just a thought.

                  --
                  Regards, Ed http://www.usenix.org.uk - http://irc.is-cool.net
                  A TCP/IP stack was the worst feature windows ever got
                  ~
                  ~
                  :wq
                • TheMiyaca
                  ... Whatever you use, exercise it in the environment you have to make sure it will work.
                  Message 8 of 22 , Nov 7, 2005
                  • 0 Attachment
                    > Although slow, I'm eventually going to look at MySQLfs. MySQL 5 has
                    > clustering of it's own, so I don't see why not to make use of it's
                    > replication and balancing. Just a thought.

                    Whatever you use, exercise it in the environment you have to make sure
                    it will work.
                  • Godwin Stewart
                    ... Hash: SHA1 On Sat, 5 Nov 2005 14:50:31 +0100, Godwin Stewart ... I m resurrecting another old thread to say that this problem has
                    Message 9 of 22 , Jan 1, 2006
                    • 0 Attachment
                      -----BEGIN PGP SIGNED MESSAGE-----
                      Hash: SHA1

                      On Sat, 5 Nov 2005 14:50:31 +0100, Godwin Stewart <gstewart@...>
                      wrote:

                      > The only thing new in the machine's configuration since yesterday is the
                      > kernel - I was running 2.6.14 instead of 2.6.13.4 (but have switched back
                      > to the older one now).

                      I'm resurrecting another old thread to say that this problem has also been
                      solved.

                      It turns out that the older drivers for my MPEG2 decompression card (the
                      ones that created havoc with ALSA and OSS emulation - Grrrrrrr) were
                      responsible for the instability. Apparently, there were known "issues" with
                      the 2.6.14.x family of kernels that have since been resolved.

                      I'm now running 2.6.14.5 and there have been no freeze-ups since I kicked
                      out the old DXR3 drivers.

                      - --
                      G. Stewart - gstewart@...

                      Your fault: core dumped
                      -----BEGIN PGP SIGNATURE-----
                      Version: GnuPG v1.4.1 (GNU/Linux)

                      iD8DBQFDt7/NK5oiGLo9AcYRAlzDAKCtOCq0lPafomJdc1mSYjK2mBtlMQCffLl2
                      asWWI/wlxr8qyVn+g0+3AtU=
                      =6G6X
                      -----END PGP SIGNATURE-----
                    Your message has been successfully submitted and would be delivered to recipients shortly.