You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Vihang Karajgaonkar (JIRA)" <ji...@apache.org> on 2017/03/04 02:23:45 UTC

[jira] [Comment Edited] (HIVE-16109) TestDbTxnManager generates a huge hive.log

    [ https://issues.apache.org/jira/browse/HIVE-16109?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15895403#comment-15895403 ] 

Vihang Karajgaonkar edited comment on HIVE-16109 at 3/4/17 2:23 AM:
--------------------------------------------------------------------

Looked into this a little deeper and realized that HIVE-13335 changed {{TxnStore.TIMED_OUT_TXN_ABORT_BATCH_SIZE}} from 1000 to 50000. The test {{TestDbTxnManager.testLockTimeout}} is using this value to create 50017 locks and waiting for some time so that they expire. This is causing the test batch to time-out since it takes very long time. The test batch timeout currently is 40 min and it takes a lot longer than that.


was (Author: vihangk1):
Looked into this a little deeper and realized that HIVE-13335 changed {{TxnStore.TIMED_OUT_TXN_ABORT_BATCH_SIZE}} from 1000 to 50000. The test {{TestDbTxnManager.testLockTimeout}} is using this value to create 50017 locks and waiting for 5 min so that they expire. This is causing the test batch to time-out since it takes very long time. The test batch timeout currently is 40 min and it takes a lot longer than that.

> TestDbTxnManager generates a huge hive.log
> ------------------------------------------
>
>                 Key: HIVE-16109
>                 URL: https://issues.apache.org/jira/browse/HIVE-16109
>             Project: Hive
>          Issue Type: Bug
>            Reporter: Vihang Karajgaonkar
>            Assignee: Vihang Karajgaonkar
>         Attachments: HIVE-16109.01.patch
>
>
> Pre-commit jobs are failing currently due to out of disk space. The issue is happening due to huge size of hive.log when TestDbTxnManager test fails or times-out. When this test fails or times-out Ptest tries to persist these logs for debugging. Since this test has been timing out frequently, this accumulates a lot of these log files and eventually Ptest server runs of disk space. Each run of TestDbTxnManager is generating ~30 GB of hive.log. I tried to run it locally and it quickly reached 7 GB until I had to cancel it.
> The issue seems to be coming from this code block in TxnHandler.java
> {noformat}
> if(LOG.isDebugEnabled()) {
>         LOG.debug("Locks to check(full): ");
>         for(LockInfo info : locks) {
>           LOG.debug("  " + info);
>         }
>       }
> {noformat}
> We should either change it trace or change the log mode of this test to INFO so that it generates smaller log files.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)