You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@cassandra.apache.org by CPC <ac...@gmail.com> on 2019/02/07 07:58:37 UTC
range repairs multiple dc
Hi All,
I searched over documentation but could not find enough reference regarding
-pr option. In some documentation it says you have to cover all ring in
some places it says you have to run it on every node regardless of you have
multiple dc.
In our case we have three dc (DC1,DC2,DC3) with every DC having 4 nodes and
12 nodes cluster in total. If i run "nodetool repair -pr --full" on very
node in DC1, does it means DC1 is consistent but DC2 and DC3 is not or DC1
is not consistent at all? Because in our case we added DC3 to our cluster
and will remove DC2 from cluster so i dont care whether DC2 have consistent
data. I dont want to repair DC2.
Also can i run "nodetool repair -pr -full" in parallel? I mean run it at
the same time in each DC or run it more than one node in same DC? Does
-dcpar option making the same thing?
Best Regards...
RE: range repairs multiple dc
Posted by Kenneth Brotman <ke...@yahoo.com.INVALID>.
This webpage has relevant information on procedures you need to use: https://docs.datastax.com/en/cassandra/3.0/cassandra/operations/opsAddDCToCluster.html
From: Kenneth Brotman [mailto:kenbrotman@yahoo.com.INVALID]
Sent: Thursday, February 07, 2019 1:31 PM
To: user@cassandra.apache.org
Subject: RE: range repairs multiple dc
A nice article on The Last Pickle blog at http://thelastpickle.com/blog/2017/12/14/should-you-use-incremental-repair.html should be helpful to you. A line in the comments following the article states:
“So restricting a -pr repair on a specific datacenter will be forbidden by Cassandra to prevent an incomplete repair from being performed.”
Give it a read.
Kenneth Brotman
From: CPC [mailto:achalil@gmail.com]
Sent: Wednesday, February 06, 2019 11:59 PM
To: user@cassandra.apache.org
Subject: range repairs multiple dc
Hi All,
I searched over documentation but could not find enough reference regarding -pr option. In some documentation it says you have to cover all ring in some places it says you have to run it on every node regardless of you have multiple dc.
In our case we have three dc (DC1,DC2,DC3) with every DC having 4 nodes and 12 nodes cluster in total. If i run "nodetool repair -pr --full" on very node in DC1, does it means DC1 is consistent but DC2 and DC3 is not or DC1 is not consistent at all? Because in our case we added DC3 to our cluster and will remove DC2 from cluster so i dont care whether DC2 have consistent data. I dont want to repair DC2.
Also can i run "nodetool repair -pr -full" in parallel? I mean run it at the same time in each DC or run it more than one node in same DC? Does -dcpar option making the same thing?
Best Regards...
RE: range repairs multiple dc
Posted by Kenneth Brotman <ke...@yahoo.com.INVALID>.
A nice article on The Last Pickle blog at http://thelastpickle.com/blog/2017/12/14/should-you-use-incremental-repair.html should be helpful to you. A line in the comments following the article states:
“So restricting a -pr repair on a specific datacenter will be forbidden by Cassandra to prevent an incomplete repair from being performed.”
Give it a read.
Kenneth Brotman
From: CPC [mailto:achalil@gmail.com]
Sent: Wednesday, February 06, 2019 11:59 PM
To: user@cassandra.apache.org
Subject: range repairs multiple dc
Hi All,
I searched over documentation but could not find enough reference regarding -pr option. In some documentation it says you have to cover all ring in some places it says you have to run it on every node regardless of you have multiple dc.
In our case we have three dc (DC1,DC2,DC3) with every DC having 4 nodes and 12 nodes cluster in total. If i run "nodetool repair -pr --full" on very node in DC1, does it means DC1 is consistent but DC2 and DC3 is not or DC1 is not consistent at all? Because in our case we added DC3 to our cluster and will remove DC2 from cluster so i dont care whether DC2 have consistent data. I dont want to repair DC2.
Also can i run "nodetool repair -pr -full" in parallel? I mean run it at the same time in each DC or run it more than one node in same DC? Does -dcpar option making the same thing?
Best Regards...