MooseFS tuning notes

The report says that the Write Performance of the moosefs group has dropped from 10 MB/sec to 3 MB/sec. (Read performance: 10 MB/sec unchanged)

Check.

Set the default one in MFSMASTER. CFG:
CHUNKS_WRITE_REP_LIMIT = 1
CHUNKS_READ_REP_LIMIT = 5

Changed:
CHUNKS_WRITE_REP_LIMIT = 0
CHUNKS_READ_REP_LIMIT = 0

The write speed is restored to 10 MB/sec.

Then change:
CHUNKS_WRITE_REP_LIMIT = 10
CHUNKS_READ_REP_LIMIT = 50

The write speed is reduced to KB/sec.

These parameters are used to adjust the intensity of moosefs balancing chunk to the new machine. The test confirms that the read and write operations of the client are not controlled.
It seems that the balanced chunk of moosefs between ChunkSever will result in a decline in the overall cluster performance. It is expected that four new machines will be added to the moosefs group a few days ago.

The official FAQ also said that this replication is indeed "it can substantially slow down the whole system .".

Http://www.moosefs.org/moosefs-faq.html#rebalancing-speed-official link

There are few parameters that can be adjusted for the performance of moosefs. Here is the only place in the configuration file that can be adjusted for the performance.

P.s. XFS used in moosefs is always dead. I am still honest and practical about ext3 .. (2011.11.10)

For details about MooseFS, click here
MooseFS: click here

Distributed File System (MooseFS)

Install and deploy the MooseFS Distributed File System on CentOS

MooseFS Distributed File System installation and configuration FAQ

MooseFS Distributed File System Installation Wizard PDF

DRBD + Heartbeat + Pacemaker achieves high availability of MFS (MooseFS)