some issues

[Thilo Raufeisen]

Some issues

Hi,
I´m testing with the mysql-cluster.
One of my tests was creating a simple table and fire masses of data in it.
After a few thousand rows, i got the "table is full" - error.
Ok, the cluster stores the data in memory, so the "DataMemory" was full. No problem to increase it.


1)
But: How can I determinate how much RAM/Space is left for the cluster, so that I´ve enough time to modify the DataMemory-variable or order a new storagenode/RAM?


2)
The next problem is speed.
I´ve a little script, that inserts 500.000 rows of data with a simple insert-statement in a loop.
On an athlon xp 3000 with 2 Gig RAM and a "normal" mysqld, this takes around 55 seconds.
The same script takes around 10 minutes on my test-cluster.
This test-cluster has 1 managment, 2 storage and 2 mysql. the storage and the mysql are on the same server. The "NoOfReplicas" is "2".

I´ve read in this forum, that using transactions and inserting multiple rows in one insert-statement would help a lot.
So I´ve modified the testscript, so it inserts 70 rows with one insert-statement. It also uses transactions and it commits each 100 insert-statement. So 7.000 new rows in a commit.
The script now takes around 70 seconds.
This seems good for me, because the clusterhardware only has an athlon xp 2600 with 1 gig ram. The script is also connecting with tcp/ip instead of socket as before.
Then I´ve added 2 additional nodes to the cluster.

Both act as storagenode. The "NoOfReplicas" is still "2".

After adding this 2 new nodes, the script doesn´t run faster.
I thought that the cluster is getting faster, when adding new nodes?


3)
adding new nodes.
What is the "perfect way" for adding new storagenodes?
I found only 1 way:
backup all data, stop the nodes, re-start the management with the new configuration, start all nodes with --initial, restore the backup


best regards