phoenix-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Taylor (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (PHOENIX-5084) Changes from Transactional Tables are not visible to query in different client
Date Tue, 01 Jan 2019 16:46:00 GMT

    [ https://issues.apache.org/jira/browse/PHOENIX-5084?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16731637#comment-16731637
] 

James Taylor commented on PHOENIX-5084:
---------------------------------------

I’d stick with v2 and make those minor changes you mentioned. We normally start the transaction
from updateCache so that we get a consistent snapshot for both data and meta data. With v3
we’d lose that plus we’d double the RPCs and transaction manager interactions.
A fourth alternative would be to pass another boolean flag through the getTableNoCache call
that would cause us to not start a transaction (and just use LATEST_TIMESTAMP) and not add
to the cache the metadata results we get back (since we wouldn’t have the proper time stamp
to use).

> Changes from Transactional Tables are not visible to query in different client
> ------------------------------------------------------------------------------
>
>                 Key: PHOENIX-5084
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-5084
>             Project: Phoenix
>          Issue Type: Bug
>    Affects Versions: 4.15.0, 4.14.1
>            Reporter: Lars Hofhansl
>            Assignee: Lars Hofhansl
>            Priority: Blocker
>         Attachments: PHOENIX-5084-v2.txt, PHOENIX-5084-v3.txt, PHOENIX-5084.txt
>
>
> Scenario:
> # Upsert and commit some data into a transactional table. (Autocommit or following by
explicit commit)
> # Query same table from another client
> The first query on the other client will not see the newly upserted/committed data (regardless
of how long one waits).
> A second identical query will see the new data.
> This happens with both Omid and Tephra.
> I guess we can't write a test for this, since it requires multiple JVMs.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message