hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Matteo Bertozzi (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-11724) Add to RWQueueRpcExecutor the ability to split get and scan handlers
Date Wed, 13 Aug 2014 18:38:16 GMT

    [ https://issues.apache.org/jira/browse/HBASE-11724?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14095893#comment-14095893
] 

Matteo Bertozzi commented on HBASE-11724:
-----------------------------------------

{quote}
Give examples in the hbase-defaults.xml for these settings – it would clarify some of the
confusion I've encountered. Give an example settings for scenarios where you have gets and
scans as equals, where gets get a lot more priority and where scans get a lot more priority.
Probably want the same with writes vs gets/scans as well.
{quote}
The example of the 0.5 value in the read.share seems to have thrown you off... 
{code}
<name>hbase.ipc.server.callqueue.read.share</name>
<description>Split the call queues into read and write queues.
      A value of 0 indicate to not split the call queues.
      A value of 0.5 means there will be the same number of read and write queues
      A value of 1.0 means that all the queues except one are used to dispatch read requests.
</description>
{code}
adding something like this? "A value > 0.5 means that there will be more read queues than
write queues" and "A value < 0.5 means that there will be less read queues than write queues".

{code}
<name>hbase.ipc.server.callqueue.scan.share</name>
<value>0</value>
<description>Used in conjunction with hbase.ipc.server.callqueue.read.share
  will split the read call queues into small-read and long-read queues.
  A value of 0 or 1 indicate to use the same set of queues for gets and scans.
</description>
{code}
Adding something like this? "A value of 0.5 means that there will be the same number of short-read
and long-read queues" and "A value > 0.5 means that there will be more long-read queues
then short-read queues" and "A value < 0.5 means that there will be less long-read queues
then short-read queues" 

{quote}
my point is that as a user you are asking me to look at the code. Let's assume I'm a clever
user and looked at the code — you haven't made any of these invariants known. comments in
the hbase-defaults or explicit checks in code would be helpful.
{quote}
Why? "share" may be interpreted as a boolean but if you see 0 as default instead of false,
it is probably not a boolean.. then you read about the 0 and 1 edge case... and you may start
understanding that that probably is the limit...
otherwise, you are just the fancy user that wants to break things and set random crap as value...
If you want I can add "the value must be between 0.0 and 1.0" to the doc... but you may want
to add that  also to all the other fields that we have in hbase-default.xml

> Add to RWQueueRpcExecutor the ability to split get and scan handlers
> --------------------------------------------------------------------
>
>                 Key: HBASE-11724
>                 URL: https://issues.apache.org/jira/browse/HBASE-11724
>             Project: HBase
>          Issue Type: New Feature
>          Components: IPC/RPC
>            Reporter: Matteo Bertozzi
>            Assignee: Matteo Bertozzi
>            Priority: Minor
>             Fix For: 2.0.0
>
>         Attachments: HBASE-11724-v0.patch
>
>
> RWQueueRpcExecutor has the devision between reads and writes requests, but we can split
also small-reads and long-reads. This can be useful to force a deprioritization of scans on
the RS.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message