List:Cluster« Previous MessageNext Message »
From:Antonio Modesto Date:July 17 2012 5:49pm
Subject:NDBD Start process
View as plain text  
Hi,


	I would like to know why sometimes the ndbd nodes take so longer to
start after a configuration change, here is what appears at the
ndb_x_out.log:

jbalock thr: 0 waiting for lock, contentions: 20 spins: 120
jbalock thr: 2 waiting for lock, contentions: 12 spins: 1271
jbalock thr: 2 waiting for lock, contentions: 13 spins: 1272
jbalock thr: 2 waiting for lock, contentions: 14 spins: 4331
jbalock thr: 2 waiting for lock, contentions: 15 spins: 4332
jbalock thr: 2 waiting for lock, contentions: 16 spins: 4333
jbalock thr: 2 waiting for lock, contentions: 17 spins: 4334
jbalock thr: 2 waiting for lock, contentions: 18 spins: 4335
jbalock thr: 2 waiting for lock, contentions: 19 spins: 5510
jbalock thr: 2 waiting for lock, contentions: 20 spins: 5511

After that it completes some phases, but this whole process took a bit
longer:

2012-07-17 14:39:22 [ndbd] INFO     -- Start phase 0 completed
2012-07-17 14:39:32 [ndbd] INFO     -- findNeighbours from: 2098 old
(left: 65535 right: 65535) new (4 4)
saving 0x8035c8000 at 0xc25980 (0)
Dbspj::execSTTOR() inst:0 phase=1
2012-07-17 14:39:32 [ndbd] INFO     -- Start phase 1 completed

jbalock thr: 0 waiting for lock, contentions: 1 spins: 1
jbalock thr: 2 waiting for lock, contentions: 1 spins: 1
jbalock thr: 0 waiting for lock, contentions: 2 spins: 53
jbalock thr: 2 waiting for lock, contentions: 2 spins: 2
jbalock thr: 0 waiting for lock, contentions: 3 spins: 73
jbalock thr: 2 waiting for lock, contentions: 3 spins: 3
jbalock thr: 2 waiting for lock, contentions: 4 spins: 4
jbalock thr: 2 waiting for lock, contentions: 5 spins: 1118
jbalock thr: 0 waiting for lock, contentions: 4 spins: 74
jbalock thr: 0 waiting for lock, contentions: 5 spins: 75
jbalock thr: 0 waiting for lock, contentions: 6 spins: 76
jbalock thr: 2 waiting for lock, contentions: 6 spins: 1119

A lot more of these messages and then:

2012-07-17 14:53:01 [ndbd] INFO     -- Start phase 2 completed
saving 0x803528000 at 0xc21730 (0)
2012-07-17 14:53:01 [ndbd] INFO     -- Start phase 3 completed
Dbspj::execSTTOR() inst:0 phase=4
2012-07-17 14:53:08 [ndbd] INFO     -- Start phase 4 completed
Using 1 fragments per node
2012-07-17 14:53:52 [ndbd] INFO     -- Start phase 5 completed
2012-07-17 14:53:52 [ndbd] INFO     -- Start phase 6 completed
m_active_buckets.set(0)
2012-07-17 14:53:52 [ndbd] INFO     -- Start phase 7 completed
2012-07-17 14:53:52 [ndbd] INFO     -- Start phase 8 completed
2012-07-17 14:53:52 [ndbd] INFO     -- Start phase 9 completed
2012-07-17 14:53:52 [ndbd] INFO     -- Start phase 100 completed
2012-07-17 14:53:52 [ndbd] INFO     -- Start phase 101 completed
2012-07-17 14:53:52 [ndbd] INFO     -- Node started


As you can see, it took more than ten minutes to start, is that normal?
The two data nodes are virtual machines, but both with two quad-core
cpu's and 8GB of RAM.


Thanks.



Thread
NDBD Start processAntonio Modesto17 Jul
  • Re: NDBD Start processWagner Bianchi17 Jul
    • Re: NDBD Start processAntonio Modesto17 Jul