accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Adam Fuchs (JIRA)" <>
Subject [jira] [Commented] (ACCUMULO-4692) CompactionDriver leaves abandoned metadata scans
Date Fri, 04 Aug 2017 17:16:00 GMT


Adam Fuchs commented on ACCUMULO-4692:

Interesting -- since I was always compacting a range with the end equalling the tablet end
row I wasn't thinking of the challenge of finding the next split point. However, I can think
of a couple of ways of getting a tighter end in the scan: 

(1) Keep the end row used to launch compactions in the previous REPO or the end row of interest
from the last time the CompactionDriver.isReady method was called in this FATE transaction.
Detect when it's wrong (i.e. after a tablet merge) and fall back to a bigger scan.
(2) Use an iterator to detect when the tablet being scanned has a prev end row after the end
of the scanned range and terminate the scan. 

Found this on a version derived from 1.7.2.

> CompactionDriver leaves abandoned metadata scans
> ------------------------------------------------
>                 Key: ACCUMULO-4692
>                 URL:
>             Project: Accumulo
>          Issue Type: Bug
>          Components: fate
>            Reporter: Adam Fuchs
> We wrote a tool to kick off tablet compactions in the background while minimizing compaction
load per-server. The tool uses range compaction on one tablet per call. We're seeing a high
number of scans on the metadata table (~7,000 on a ~100 node cluster).
> The metadata query in the isReady() method of CompactionDriver that is used to see if
the compaction has completed uses a range that goes to the end of the metadata entries for
the given table, but it stops consuming the results of the scanner at the end of the compaction
range. isReady gets called in a pretty tight loop, especially with hundreds of compactions
running concurrently. Seems like we should limit the scan to the metadata range associated
with the compaction so that the scan can get cleaned up quickly.

This message was sent by Atlassian JIRA

View raw message