[
https://issues.apache.org/jira/browse/CASSANDRA-2081?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12989828#comment-12989828
]
Aaron Morton commented on CASSANDRA-2081:
-----------------------------------------
Looking at the right code now.
In StorageProxy.fetchRows() line 395 the digestMessage is shared for all non
local digest requests.
And I just realised the log messages above are coming from
StorageProxy.getRangeSlice() where the message object is shared for all
endpoints. The log messages from Thibaut show the same message ID on nodes 0.1
0.2 and 0.3.
My reading of the MessagingService and ResponseVerbHandler is that once a
message with ID X has been received, if others are received with the same ID
they will be ignored. Is that correct? If so this looks like a bug in the two
methods above.
> Consistency QUORUM does not work anymore (hector:Could not fullfill request
> on this host)
> -----------------------------------------------------------------------------------------
>
> Key: CASSANDRA-2081
> URL: https://issues.apache.org/jira/browse/CASSANDRA-2081
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Environment: linux, hector + cassandra
> Reporter: Thibaut
> Priority: Blocker
> Fix For: 0.7.1
>
> Attachments: 2081-logging.patch
>
>
> I'm using apache-cassandra-2011-01-28_20-06-01.jar and hector 7.0.25.
> Using consistency level Quorum won't work anymore (tested it on read).
> Consisteny level ONE still works though
> I have tried this with one dead node in my cluster.
> If I restart cassandra with an older svn revision
> (apache-cassandra-2011-01-28_20-06-01.jar), I can access the cluster with
> consistency level QUORUM again, while still using
> apache-cassandra-2011-01-28_20-06-01.jar and hector 7.0.25 in my application.
> 11/01/31 19:54:38 ERROR connection.CassandraHostRetryService: Downed
> intr1n18(192.168.0.18):9160 host still appears to be down: Unable to open
> transport to intr1n18(192.168.0.18):9160 , java.net.NoRouteToHostException:
> No route to host
> 11/01/31 19:54:38 INFO connection.CassandraHostRetryService: Downed Host
> retry status false with host: intr1n18(192.168.0.18):9160
> 11/01/31 19:54:45 ERROR connection.HConnectionManager: Could not fullfill
> request on this host CassandraClient<intr1n11:9160-483>
> intr1n11 is marked as up however and I can also access the node through the
> cassandra cli.
> 192.168.0.1 Up Normal 8.02 GB 5.00% 0cc
> 192.168.0.2 Up Normal 7.96 GB 5.00% 199
> 192.168.0.3 Up Normal 8.24 GB 5.00% 266
> 192.168.0.4 Up Normal 4.94 GB 5.00% 333
> 192.168.0.5 Up Normal 5.02 GB 5.00% 400
> 192.168.0.6 Up Normal 5 GB 5.00% 4cc
> 192.168.0.7 Up Normal 5.1 GB 5.00% 599
> 192.168.0.8 Up Normal 5.07 GB 5.00% 666
> 192.168.0.9 Up Normal 4.78 GB 5.00% 733
> 192.168.0.10 Up Normal 4.34 GB 5.00% 7ff
> 192.168.0.11 Up Normal 5.01 GB 5.00% 8cc
> 192.168.0.12 Up Normal 5.31 GB 5.00% 999
> 192.168.0.13 Up Normal 5.56 GB 5.00% a66
> 192.168.0.14 Up Normal 5.82 GB 5.00% b33
> 192.168.0.15 Up Normal 5.57 GB 5.00% c00
> 192.168.0.16 Up Normal 5.03 GB 5.00% ccc
> 192.168.0.17 Up Normal 4.77 GB 5.00% d99
> 192.168.0.18 Down Normal ? 5.00% e66
> 192.168.0.19 Up Normal 4.78 GB 5.00% f33
> 192.168.0.20 Up Normal 4.83 GB 5.00% ffffffffffffffff
--
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira
