From: Andrew Morgan
Date: December 18 2012 5:47pm
Subject: Re: MySQL Cluster alerts
List-Archive: http://lists.mysql.com/mysql/228828
Message-Id: <61cdc84f-b6fb-4248-ba79-6928cd15c531@default>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: quoted-printable

Hi Bheemsen,

 looks like a few different things going on there; if you have a MySQL supp=
ort contract/subscription then it would be worth raising SRs - it doesn't n=
eed to be a bug, it's fine tyo ask questions too. A couple of things that s=
pring to mind in-line....

> I am frequently seeing the following alerts in our production MySQL Clust=
er
> environment. Do you have any metrics, guidelines and scripts to monitor a=
nd
> fix these alerts? Any help is appreciated.
>
> Temporary Tables To Disk Ratio Excessive
> Excessive Disk Temporary Table Usage Detected
>
> Table Scans Excessive
> Indexes Not Being Used Efficiently

If you're using MySQL Cluster 7.2 then you should run OPTIMIZE TABLE for ea=
ch of your tables (repeat that step whenever you make schemas changes to it=
, add an index or make very signifficant data changes). This will make the =
optimizer make better use of available indexes. Use the query analyzer in M=
ySQL Enterprise Monitor (MEM) to see which queries are taking the time as t=
hese are likely to be the table scans (full table scans should be avoided a=
s much as possible). You can use the EXPLAIN command to see if individual q=
ueries are making use of the available indexes. Try adding new indexes if t=
hey're missing for high-running transactions.

>
> Thread Cache Size May Not Be Optimal
>
> Cluster DiskPageBuffer Hit Ratio Is Low

Note that you might observe this after restarting a data node as the cache =
must be repopulated as queries come in. If you're seeing this at other time=
s or the MEM graphs show that the DiskPageBuffer Hit Ratio is consistently =
low then consider increasing it... http://dev.mysql.com/doc/mysql-cluster-e=
xcerpt/5.1/en/mysql-cluster-ndbd-definition.html#ndbparam-ndbd-diskpagebuff=
ermemory=20