GET 24/7 LIVE HELP NOW

Announcement

Announcement Module
Collapse
No announcement yet.

Enable Flashcache in 5.5

Page Title Module
Move Remove Collapse
X
Conversation Detail Module
Collapse
  • Filter
  • Time
  • Show
Clear All
new posts

  • Enable Flashcache in 5.5

    I'm running multiple servers with Percona/MySQL 5.1.16 and Flashcache, which works quite fine.

    I'm now however having problems installing Percona/MySQL 5.5.19 on new servers with Flashcache, as the startup states there is a problem with the "Flashcache bypass".

    Output from a server running 5.5.19:

    120118 03:51:25 mysqld_safe Starting mysqld daemon with databases from /var/lib/mysql120118 3:51:25 [Note] Flashcache bypass: disabled120118 3:51:25 [Note] Flashcache setup error is : ioctl failed120118 3:51:25 [Note] Plugin 'FEDERATED' is disabled.120118 3:51:25 InnoDB: The InnoDB memory heap is disabled120118 3:51:25 InnoDB: Mutexes and rw_locks use GCC atomic builtins120118 3:51:25 InnoDB: Compressed tables use zlib 1.2.3120118 3:51:25 InnoDB: Using Linux native AIO120118 3:51:25 InnoDB: Initializing buffer pool, size = 120.0G120118 3:51:31 InnoDB: Completed initialization of buffer pool120118 3:51:31 InnoDB: highest supported file format is Barracuda.120118 3:51:36 InnoDB: Waiting for the background threads to start120118 3:51:37 Percona XtraDB (http://www.percona.com) 1.1.8-rel24.0 started; log sequence number 324537592488120118 3:51:37 [Note] Slave SQL thread initialized, starting replication in log 'mysql-bin.000104' at position 74933008, relay log '/database/dblogs/mysql-relay-bin.000272' position: 67383397120118 3:51:37 [Note] Event Scheduler: Loaded 0 events120118 3:51:37 [Note] /usr/sbin/mysqld: ready for connections.Version: '5.5.19-55-log' socket: '/var/lib/mysql/mysql.sock' port: 3306 Percona Server (GPL), Release rel24.0, Revision 204


    If I however downgrade to Percona/MySQL 5.1.16 on the same server Flashcache works:

    120118 05:48:12 mysqld_safe Starting mysqld daemon with databases from /var/lib/mysql120118 5:48:12 [Note] Flashcache bypass: enabled120118 5:48:12 [Note] Plugin 'FEDERATED' is disabled.InnoDB: The InnoDB memory heap is disabledInnoDB: Mutexes and rw_locks use GCC atomic builtinsInnoDB: Compressed tables use zlib 1.2.3120118 5:48:12 InnoDB: Initializing buffer pool, size = 120.0G120118 5:48:17 InnoDB: Completed initialization of buffer pool120118 5:48:17 InnoDB: highest supported file format is Barracuda.120118 5:48:22 Percona XtraDB (http://www.percona.com) 1.0.17-rel13.1 started; log sequence number 45356120118 5:48:22 [Note] Event Scheduler: Loaded 0 events120118 5:48:22 [Note] /usr/sbin/mysqld: ready for connections.Version: '5.1.60-rel13.1-log' socket: '/var/lib/mysql/mysql.sock' port: 3306 Percona Server (GPL), 13.1, Revision 413


    Flashcache is mounted on /var/lib/mysql.


    Is there a problem with Flashcache in Percona/MySQL 5.5, or does it wrongfully report that Flashcache bypass can't be enabled?
    Is there any way for me to validate that MySQL works correctly, besides seeing that our SSD harddrives work as expected?

  • #2
    I see the exact same thing with Percona 5.5.20 also. It would be nice to find out how to resolve this with out having to run 5.1.

    Comment


    • #3
      I see no one else has responded and frankly I'm not sure -- I'm not the expert in administering Flashcache. So you're likely to get best results by engaging us for consulting.

      Comment


      • #4
        Thanks for replying Baron - we're already in progress of receiving consulting from Percona on this issue.

        Comment


        • #5
          Engaging in consulting for help on an item that if I make no other change on my server but run Percona 5.1 vs. 5.5 and I get that message sounds like a software problem to me and not a flashcache config issue. I do not see why I should pay for support to be told why there is a problem with Percona.

          Comment


          • #6
            Why would you pay for support on anything? You think that just because it's free software, all your bugs will be diagnosed and fixed for free? I'm not sure what I can say to that.

            Comment


            • #7
              I have no problem paying for support, but when it is obviously a bug then I do not think I need to pay to be told it is a bug. All someone had to do was reply to the thread that it was a bug, a change in the code, something that lets me know Percona was aware of it. I do not even care if you admited to it being fixed sooner or later. But to reply to the question with you are "likely to get best results by engaging us for consulting" to get the answer. That just tells me Percona only cares about the people paying for support and the forum really is for community support so do not expect an answer from Percona unless you pay for it.

              Comment


              • #8
                I see -- I misread your comment. Sorry for jumping to conclusions.

                You misread mine, too, though. I meant that obviously everyone who's competent to help you (it's not me) is busy, so if you want attention, you probably have to pay for it.

                I actually care a lot about helping people out when I can. I make it a point to go though all of the forum questions once a week and reply to them if I haven't yet, and I watch the RSS feed and reply quickly when I can. But some things require some thought, so I wait until I can dedicate some time to it (task-switching is a productivity killer), and if I can't help resolve the issue then, as in this case, I recommend paying for more help. If you look around the forum, I answered a number of questions yesterday in my weekly run-through, some of them non-trivial, which I could have easily said "you gotta pay for that answer." So "Percona only cares about the people paying for support" might be how it looks but it's not really true. Heck, if we only cared about people who pay, we wouldn't release so much free software, much of which I've personally spent evenings and weekends developing over the last 6 years.

                Anyway, yes it's true this is just a community support forum on a best-effort basis, and a paid escalation is always an option to recommend, but my point is we're not here just trying to twist everyone's arm into paying for help.

                Comment


                • #9
                  got the same problem with 5.5.17+

                  anyone found a solution for this except downgrading to 5.1?

                  Comment


                  • #10
                    found something here: https://bugs.launchpad.net/percona-server/+bug/747032

                    adding

                    cd $datapath

                    into /etc/init.d/mysql fixes this problem.

                    cheers,

                    Juergen

                    Comment


                    • #11
                      Hi Juergen,

                      where exactly did you add that line in order to fix the issue? I've tried to add it, too, but the issue persists. (Please post the precise snippet!)

                      Thanks!
                      - Denis

                      Comment


                      • #12
                        i dropped it to line 283

                        # Give extra arguments to mysqld with the my.cnf file. This script
                        # may be overwritten at next upgrade.
                        cd $datadir
                        $bindir/mysqld_safe --datadir="$datadir" --pid-file="$mysqld_pid_fil

                        cheers,

                        Juergen

                        Comment


                        • #13
                          Thanks Jürgen! That doesn't do the trick for me unfortunately, though. We're definitely using different OS-s, my mysql init script doesn't even have that many lines... I inserted it before the call to mysqld_safe, which does not help.
                          (I'm on Ubuntu 10.04, Percona Server 5.5.21.)

                          Comment

                          Working...
                          X