[ 
https://issues.apache.org/jira/browse/HBASE-14938?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15071505#comment-15071505
 ] 

Ashish Singhi commented on HBASE-14938:
---------------------------------------

Thanks for the comments.

[~anoop.hbase]
bq. Can we avoid having 2 loop code.. When totalNoOfFiles <maxZnodesPerRequest, 
I initially thought u will get issue in 1st loop.. But then totalNoOfRequests 
will be 0 and so we wont go into 1st loop at all.. I would say readability wise 
this can be improved.
I had that in my first patch, then as per Ted suggestion for better readability 
I combined them. Anyways when files are less than the max value we will have to 
do a if condition check to avoid the first loop, that condition check we have 
now in for loop, is that bad ?

bq. totalNoOfRequests = 1 in that case and every case we have one loop doing 
the work?
No, for example when we have totalNoOfFiles=6000 we will execute both the loops.

bq. How we have selected the def value to be 5000? We will get these many files 
to be added in one call?
Mostly not, this is only to control the max data size limit allowed in a ZK 
request (as mentioned in description).

[~tedyu]

bq. When would totalNoOfRequests be 1 ?
When totalNoOfFiles is equal to maxZnodesPerRequest.

> Limit the number of znodes for ZK in bulk loaded hfile replication
> ------------------------------------------------------------------
>
>                 Key: HBASE-14938
>                 URL: https://issues.apache.org/jira/browse/HBASE-14938
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: Ashish Singhi
>            Assignee: Ashish Singhi
>             Fix For: 2.0.0, 1.3.0
>
>         Attachments: HBASE-14938(1).patch, HBASE-14938-v1.patch, 
> HBASE-14938.patch
>
>
> In ZK the maximum allowable size of the data array is 1 MB. Until we have 
> fixed HBASE-10295 we need to handle this.
> Approach to this problem will be discussed in the comments section.
> Note: We have done internal testing with more than 3k nodes in ZK yet to be 
> replicated.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to