Bug #11131 all fragments of the System tables are stored on one node
Submitted: 7 Jun 2005 9:30 Modified: 6 Jul 2005 11:30
Reporter: Jan Kneschke Email Updates:
Status: Won't fix Impact on me:
None 
Category:MySQL Cluster: Cluster (NDB) storage engine Severity:S3 (Non-critical)
Version:4.1.12 OS:Linux (Linux/EM64T (RHEL 4))
Assigned to: Assigned Account CPU Architecture:Any

[7 Jun 2005 9:30] Jan Kneschke
Description:
All fragments of the NDB_* and SYS tables are stored on one Node in the cluster. 

How to repeat:
Do a backup of the cluster and use ndb_restore to replay it into the cluster.

The fragments of the system table will only be dumped for the first node.

Suggested fix:
Distribute the fragements over all nodes.
[7 Jun 2005 9:38] Jan Kneschke
Added version information
[8 Jun 2005 5:23] Jorge del Conde
Tested w/4.1 from bk
[6 Jul 2005 11:12] Pekka Nousiainen
All tables (including system tables) are stored on all nodes.

Assume two db nodes 1 and 2 with NoOfReplicas=2.  By default a table
has one fragment per node.   A fragment has 2 replicas: primary (P) and
backup (B).   Currently a table gets following fragments:

node 1:     P   B
node 2:     B   P

The thing noticed in this bug# is a defiency in the fragment calculation.
First 2 tables created get these fragments:

node 1:    P   P
node 2:    B   B

Fixing this could slightly improve autoincrement performance.

Another minor defiency is that the fragments should alternate
for each new table between these 2:

node 1:     P   B
node 2:     B   P

node 1:     B   P
node 2:     P   B
[13 Mar 2014 13:33] Omer Barnir
This bug is not scheduled to be fixed at this time.