RE: NDBCluster: Table is full - Crouch, Luke H. - com.mysql.lists.cluster

4 messages in com.mysql.lists.clusterRE: NDBCluster: Table is full

From	Sent On	Attachments
kasandra Padisha	15 Sep 2004 14:19
Crouch, Luke H.	15 Sep 2004 14:45
Tomas Ulin	16 Sep 2004 02:21
Crouch, Luke H.	16 Sep 2004 06:06

Subject:	RE: NDBCluster: Table is full
From:	Crouch, Luke H. (Luke...@red-man.com)
Date:	09/16/2004 06:06:49 AM
List:	com.mysql.lists.cluster

thanks for the info...it's good to know. I just remembered when we tried to load
a large amount of data with the default memory limit...after it hit the table
full error, we changed the config and tried to shutdown and then start the
cluster back up and had big problems with the cluster using about a gig of swap!
we didn't try to "regular" config change, though.

let me know if the performing this regular config refresh works for you
kasandra, I'd be very interested to find it if we could have avoided re-building
the cluster by doing things correctly! =)

thanks, -L

-----Original Message----- From: Tomas Ulin [mailto:tom...@mysql.com] Sent: Thursday, September 16, 2004 4:22 AM To: Crouch, Luke H.; kasandra Padisha Cc: clus...@lists.mysql.com Subject: Re: NDBCluster: Table is full

This is a type of config change one can do online if one is careful, following the "regular" schema.

1. restart ndb_mgmd with new config 2. restart the nodes one by one by issuing in ndb_mgm: <id1> restart <id2> restart

The upgrade is safe to do online because the memory sizes are increased (decreasing is a trickier story),

And since these changes do not affect anything regarding the filesystem, no -i flag is neccessary. The only thing that the -i flag does is to clear and reinitialize the filesystem.

Crouch, Luke H. wrote:

you need to increase your DataMemory and IndexMemory parameters. the defaults are 80M and 24M, respectively...

in your config.ini...

[DB DEFAULT] DataMemory: 500M IndexMemory: 100M

when you do this, however...I think you may want to start

the nodes with -i option. and re-load the data again. Mikael can tell you for sure, though.

-L

-----Original Message----- From: kasandra Padisha [mailto:kasa...@operamail.com] Sent: Wednesday, September 15, 2004 4:20 PM To: clus...@lists.mysql.com Subject: NDBCluster: Table is full

I have a cluster running using mysql-4.1.4a-gamma, on debians 3.0 stable. It has a master (MGM), two DBs and two (API). I included the config.ini file at the end.

It works pretty well, we have test it all, from one API, from the other, the sync works OK, .. everything looking good

..but now we are starting to load it with a big amount of information and we start to have the message:

ERROR 1114 at line 2: The table 'aaaaa' is full

We are running an script that loads lot of information, and in some places it stops with this message, and it is not always. We tried to do it by hand, with the tables in trouble, and it lets to insert one row at a time, but with two it did not.

We tried to increase the max rows as it is shown in the manual Sect A.2.11 (of course it is for myisam but we gave it a go) .. but we got an error

ALTER table aaa MAX_ROWS=10000000 AVG_ROW_LENGTH=1039335076;

and returned ERROR 1005 (HY000): Can't create table './cdr/#sql-4a98_10f.frm' (errno: 4336)

strange .. ... Any clue somebody ???

--------------------------------------- config.ini --------------- [DB DEFAULT] NoOfReplicas: 2

[COMPUTER] Id: 1 HostName: mgm

[COMPUTER] Id: 2 HostName: proc1

[COMPUTER] Id: 3 HostName: proc2

[MGM] Id: 1 ExecuteOnComputer: 1

[DB] Id: 2 ExecuteOnComputer: 2 FileSystemPath: /var/mysql/ndb/data

[DB] Id: 3 ExecuteOnComputer: 3 FileSystemPath: /var/mysql/ndb/data

[API] Id: 4 ExecuteOnComputer: 2

[API] Id: 5 ExecuteOnComputer: 3

[TCP DEFAULT] PortNumber: 10002

---------------------------------------------------

_____________________________________________________________ Web-based SMS services available at http://www.operamail.com. From your mailbox to local or overseas cell phones.