FAQ
Hi group,

Am I reading this error correctly? I can't make schema changes while one node is restarting? Is this planned to change before the final release?

The cluster cant be inaccessible for eg. creating tables just because a node is restarting. Also, due to Bug #25710 where my nodes require an --initial, the nodes take a few minutes to startup each (1GB of tablesapce. We will need about 1TB per machine for production, which would take hours/days!).


mysql> show warnings;
+-------+------+-------------------------------------------------------------------------------------------------+
Level | Code | Message |
+-------+------+-------------------------------------------------------------------------------------------------+
Error | 1297 | Got temporary error 711 'System busy with node restart, schema operations not allowed' from NDB |
Error | 1005 | Can't create table '1_2.brian' (errno: 711) |
+-------+------+-------------------------------------------------------------------------------------------------+


Brian




__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Search Discussions

  • Pekka at Jan 26, 2007 at 9:46 pm

    Am I reading this error correctly? I can't make schema changes while one node is restarting? Correct.
    Is this planned to change before the final release?
    No. Unless there is huge demand..
    The cluster cant be inaccessible for eg. creating tables just because a node is restarting.
    It is inaccessible for metadata ops.

    How often do you change the schema?

    --
    Pekka Nousiainen, Software Engineer
    MySQL AB, www.mysql.X.com
    pekka@mysql.x.com +46 (0) 73 068 4978
  • Brian at Jan 26, 2007 at 9:56 pm
    We have thousands of create and drop tables daily.

    We absolutely need to have the cluster available for schema changes through a node failure. We wanted to move to a cluster to be able to roll through node restarts without affecting clients.

    I'm testing a 4 server mysql cluster to take over the work 2 mysql 4.1
    servers using innodb are doing. We have over 10k tables, 280gb data,
    and per month: about 2 billion inserts, 2 billion updates, 600 million
    reads.


    Heh, between my first message and this one, management has already questioned the reasoning for using the cluster at all. Ack. I really don't want to go down the path of attampting application splitting to handle our write-bound data SQL load.

    Brian


    ----- Original Message ----
    From: "pekka@mysql.com" <pekka@mysql.com>
    To: cluster@lists.mysql.com
    Sent: Friday, January 26, 2007 4:51:47 PM
    Subject: Re: Error 711 System busy with node restart
    Am I reading this error correctly? I can't make schema changes while one node is restarting? Correct.
    Is this planned to change before the final release?
    No. Unless there is huge demand..
    The cluster cant be inaccessible for eg. creating tables just because a node is restarting.
    It is inaccessible for metadata ops.

    How often do you change the schema?

    --
    Pekka Nousiainen, Software Engineer
    MySQL AB, www.mysql.X.com
    pekka@mysql.x.com +46 (0) 73 068 4978

    --
    MySQL Cluster Mailing List
    For list archives: http://lists.mysql.com/cluster
    To unsubscribe: http://lists.mysql.com/cluster?unsub=spatuality@yahoo.ca







    __________________________________________________
    Do You Yahoo!?
    Tired of spam? Yahoo! Mail has the best spam protection around
    http://mail.yahoo.com
  • Pekka at Jan 26, 2007 at 10:08 pm
    We have thousands of create and drop tables daily.
    This seems strange.

    Are these temp/work tables? Then maybe they can be created
    as isam or memory tables.

    If not, can you use one big table instead of thousands?
    The big table would have an indexed key which says what data
    set (previously in separate table) in accessed.

    --
    Pekka Nousiainen, Software Engineer
    MySQL AB, www.mysql.X.com
    pekka@mysql.x.com +46 (0) 73 068 4978
  • Kevin Burton at Jan 27, 2007 at 12:12 am
    Yeah...... I have to assert that for reporting needs this can be a common
    use case. I personally try to avoid it in automated code but crontab code
    can certainly need to create tables for reporting.

    The main problem (which could solve both issues) is that TRUNCATE TABLE in
    NDB is basically just a CREATE TABLE (from what I understand) and so one
    wouldn't be able to create the tables ahead of time and just TRUNCATE them
    because TRUNCATE isn't available during node restart either.
    On 1/26/07, pekka@mysql.com wrote:

    We have thousands of create and drop tables daily.
    This seems strange.

    Are these temp/work tables? Then maybe they can be created
    as isam or memory tables.

    If not, can you use one big table instead of thousands?
    The big table would have an indexed key which says what data
    set (previously in separate table) in accessed.

    --
    Pekka Nousiainen, Software Engineer
    MySQL AB, www.mysql.X.com
    pekka@mysql.x.com +46 (0) 73 068 4978

    --
    MySQL Cluster Mailing List
    For list archives: http://lists.mysql.com/cluster
    To unsubscribe:
    http://lists.mysql.com/cluster?unsub=burtonator@gmail.com

    --
    Founder/CEO Tailrank.com
    Location: San Francisco, CA
    AIM/YIM: sfburtonator
    Skype: burtonator
    Blog: feedblog.org
    Cell: 415-637-8078

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupcluster @
categoriesmysql
postedJan 26, '07 at 7:53p
activeJan 27, '07 at 12:12a
posts5
users3
websitemysql.com
irc#mysql

3 users in discussion

Brian: 2 posts Pekka: 2 posts Kevin Burton: 1 post

People

Translate

site design / logo © 2018 Grokbase