7.2 RPM repackaging thoughts ...

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

7.2 RPM repackaging thoughts ...

Hartmut Holzgraefe-4
Starting with MySQL Cluster 7.2 there don't seem to be separate
'storage' and 'management' RPMs anymore; 'ndbd', 'ndbmtd' and
'ndb_mgmd' binaries are now part of the 'server' RPM along with
'mysqld'.

As the upgrade order for nodes during a rolling upgrade is
"management nodes, then data nodes, and only in the last step
mysqld/API nodes" I see the following problem with this new
combined RPMs:

* it is not uncommon to run ndb_mgmd or ndbd and mysqld on the
  same machines in parallel,

* to do a rolling upgrade i need to replace ndb_mgmd binaries
  first, as these are not in a separate package anymore I'm
  also forced to replace mysqld binaries now

* same for data node binaries ...

* so if during a rolling restart, which can take quite some
  time depending on DataMemory size and other factors, a
  mysqld instance fails for whatever reason and gets restarted
  (e.g. by mysqld_safe) it will use the newer mysqld version
  now even though it shouldn't by the required upgrade order

Is this something to be worried about, or is it actually not
an issue if a mysql node gets upgraded before the rolling
upgrade of data nodes has finished?

--
Hartmut Holzgraefe, Principal Support Engineer (EMEA)
MariaDB Corporation | http://www.mariadb.com/

--
MySQL Cluster Mailing List
For list archives: http://lists.mysql.com/cluster
To unsubscribe:    http://lists.mysql.com/cluster

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: 7.2 RPM repackaging thoughts ...

Jesper Wisborg Krogh
Hi Hartmut,

On 13/12/2014 02:50, Hartmut Holzgraefe wrote:

> Starting with MySQL Cluster 7.2 there don't seem to be separate
> 'storage' and 'management' RPMs anymore; 'ndbd', 'ndbmtd' and
> 'ndb_mgmd' binaries are now part of the 'server' RPM along with
> 'mysqld'.
>
> As the upgrade order for nodes during a rolling upgrade is
> "management nodes, then data nodes, and only in the last step
> mysqld/API nodes" I see the following problem with this new
> combined RPMs:
>
> * it is not uncommon to run ndb_mgmd or ndbd and mysqld on the
>    same machines in parallel,
>
> * to do a rolling upgrade i need to replace ndb_mgmd binaries
>    first, as these are not in a separate package anymore I'm
>    also forced to replace mysqld binaries now
>
> * same for data node binaries ...
>
> * so if during a rolling restart, which can take quite some
>    time depending on DataMemory size and other factors, a
>    mysqld instance fails for whatever reason and gets restarted
>    (e.g. by mysqld_safe) it will use the newer mysqld version
>    now even though it shouldn't by the required upgrade order
>
> Is this something to be worried about, or is it actually not
> an issue if a mysql node gets upgraded before the rolling
> upgrade of data nodes has finished?
>

You can upgrade the mysqld/API nodes before or after the other node
types (including the management nodes) as you wish -
https://dev.mysql.com/doc/refman/5.6/en/mysql-cluster-rolling-restart.html:

> MySQL Cluster supports a flexible order for upgrading nodes. When
> upgrading a MySQL Cluster, you may upgrade API nodes (including SQL
> nodes) before upgrading the management nodes, data nodes, or both. In
> other words, you are permitted to upgrade the API and SQL nodes in any
> order. This is subject to the following provisions:
> ...


Best regards,
Jesper Krogh
MySQL Support

--
MySQL Cluster Mailing List
For list archives: http://lists.mysql.com/cluster
To unsubscribe:    http://lists.mysql.com/cluster

Loading...