cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "J.B. Langston (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (CASSANDRA-6548) Order nodetool ring output by token when vnodes aren't in use
Date Fri, 03 Jan 2014 16:45:50 GMT

     [ https://issues.apache.org/jira/browse/CASSANDRA-6548?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

J.B. Langston updated CASSANDRA-6548:
-------------------------------------

    Description: 
It is confusing to order the nodes by hostId in nodetool ring when vnodes aren't in use. This
happens in 1.2 when providing a keyspace name:

{code}
Datacenter: DC1
==========
Replicas: 2

Address         Rack        Status State   Load            Owns                Token
                                                                               42535295865117307932921825928971026432
xxx.xxx.xxx.48   RAC2        Up     Normal  324.26 GB       25.00%              85070591730234615865843651857942052864
xxx.xxx.xxx.42   RAC1        Up     Normal  284.39 GB       25.00%              0
xxx.xxx.xxx.44   RAC1        Up     Normal  931.07 GB       75.00%              127605887595351923798765477786913079296
xxx.xxx.xxx.46   RAC2        Up     Normal  881.93 GB       75.00%              42535295865117307932921825928971026432

Datacenter: DC2
==========
Replicas: 2

Address         Rack        Status State   Load            Owns                Token
                                                                               148873535527910577765226390751398592512
xxx.xxx.xxx.19  RAC2        Up     Normal  568.22 GB       50.00%              63802943797675961899382738893456539648
xxx.xxx.xxx.17  RAC1        Up     Normal  621.58 GB       50.00%              106338239662793269832304564822427566080
xxx.xxx.xxx.15  RAC1        Up     Normal  566.99 GB       50.00%              21267647932558653966460912964485513216
xxx.xxx.xxx.21  RAC2        Up     Normal  619.41 GB       50.00%              148873535527910577765226390751398592512
{code}

Among other things, this makes it hard to spot rack imbalances.  In the above output, the
racks in DC1 are actually incorrectly ordered and those in DC2 are correctly ordered, but
it's not obvious until you manually sort the nodes by token.

  was:
It is confusing to order the nodes by hostId in nodetool ring when vnodes aren't in use. This
happens in 1.2 when providing a keyspace name:

{code}
Datacenter: DC1
==========
Replicas: 2

Address         Rack        Status State   Load            Owns                Token
                                                                               42535295865117307932921825928971026432
xxx.xxx.xxx.48   RAC2        Up     Normal  324.26 GB       25.00%              85070591730234615865843651857942052864
xxx.xxx.xxx.42   RAC1        Up     Normal  284.39 GB       25.00%              0
xxx.xxx.xxx.44   RAC1        Up     Normal  931.07 GB       75.00%              127605887595351923798765477786913079296
xxx.xxx.xxx.46   RAC2        Up     Normal  881.93 GB       75.00%              42535295865117307932921825928971026432

Datacenter: DC2
==========
Replicas: 2

Address         Rack        Status State   Load            Owns                Token
                                                                               148873535527910577765226390751398592512
xxx.xxx.xxx.19  RAC2        Up     Normal  568.22 GB       50.00%              63802943797675961899382738893456539648
xxx.xxx.xxx.17  RAC1        Up     Normal  621.58 GB       50.00%              106338239662793269832304564822427566080
xxx.xxx.xxx.15  RAC1        Up     Normal  566.99 GB       50.00%              21267647932558653966460912964485513216
xxx.xxx.xxx.21  RAC2        Up     Normal  619.41 GB       50.00%              148873535527910577765226390751398592512
{code}

Among other things, it makes it hard to spot rack imbalances.  In the above output, the racks
DC1 is actually incorrectly ordered and DC2 is correctly ordered, but it's not obvious until
you manually sort the nodes by token.


> Order nodetool ring output by token when vnodes aren't in use
> -------------------------------------------------------------
>
>                 Key: CASSANDRA-6548
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6548
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: J.B. Langston
>
> It is confusing to order the nodes by hostId in nodetool ring when vnodes aren't in use.
This happens in 1.2 when providing a keyspace name:
> {code}
> Datacenter: DC1
> ==========
> Replicas: 2
> Address         Rack        Status State   Load            Owns                Token
>                                                                                42535295865117307932921825928971026432
> xxx.xxx.xxx.48   RAC2        Up     Normal  324.26 GB       25.00%              85070591730234615865843651857942052864
> xxx.xxx.xxx.42   RAC1        Up     Normal  284.39 GB       25.00%              0
> xxx.xxx.xxx.44   RAC1        Up     Normal  931.07 GB       75.00%              127605887595351923798765477786913079296
> xxx.xxx.xxx.46   RAC2        Up     Normal  881.93 GB       75.00%              42535295865117307932921825928971026432
> Datacenter: DC2
> ==========
> Replicas: 2
> Address         Rack        Status State   Load            Owns                Token
>                                                                                148873535527910577765226390751398592512
> xxx.xxx.xxx.19  RAC2        Up     Normal  568.22 GB       50.00%              63802943797675961899382738893456539648
> xxx.xxx.xxx.17  RAC1        Up     Normal  621.58 GB       50.00%              106338239662793269832304564822427566080
> xxx.xxx.xxx.15  RAC1        Up     Normal  566.99 GB       50.00%              21267647932558653966460912964485513216
> xxx.xxx.xxx.21  RAC2        Up     Normal  619.41 GB       50.00%              148873535527910577765226390751398592512
> {code}
> Among other things, this makes it hard to spot rack imbalances.  In the above output,
the racks in DC1 are actually incorrectly ordered and those in DC2 are correctly ordered,
but it's not obvious until you manually sort the nodes by token.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)

Mime
View raw message