How long does it take to restore one crashed hbase RegionServer

It seems that the Hbase RegionServer is the only node, and not like Cassandra, there are replicable nodes. I know that when one RegionServer is turned off, HMaster will assign regions that are on the RS fault to another RS.

but how long will the new RegionServer be ready to serve the broken regions. If it took too long, the client cannot wait too long and the client will exclude even lost data, please?

+4
source share
2 answers

, , HBase .
. :

Hbase

, hbase, ...
3 :

  • , . , Heatbeat, Zookeeper. , , regionServer .
  • : . , , node, . , , , , .
  • : hbase.

?

, . , hbase zookeeper. . .

FengWang, , HBase . Cassandra, node . CAP: Hbase master/regionServer , strong > . , Cassandra , .

, HBase (, FengWang), . 100 , 10 , . , , ( , , ). Cassandra . , .

. .

+4

100 Hbase. RegionServer down Hbase 3-5 , Hlog HDFS. i.e 5 . , , 1 . 1 , , Hbase.

Cassandra, node , 1 .

0

All Articles