[jira] [Comment Edited] (IGNITE-15718) Issues in implementation of the method TableManager#table(String)

["Vladislav Pyatkov (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IGNITE-15718?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17456354#comment-17456354 ] 

Vladislav Pyatkov edited comment on IGNITE-15718 at 12/9/21, 11:31 AM:
-----------------------------------------------------------------------

Hi [~maliev]
We want to avoid returning a stale table to the client side, but it is still possible:
{code}
Table tbl = tables.get(name);
if (tbl != null) {
    return CompletableFuture.completedFuture(tbl);
}
{code}
Required to get actual ID from Metastorage before getting a table.
Also, I sure you can to write a test, where Metastorage majority create, drop and create again table with the similar name,  but one Ignite node lagging and has received yet this updates.
Use {{WatchListenerInhibitor}} to simulate a slow node.


was (Author: v.pyatkov):
Hi [~maliev]
We want to avoid returning a stale table to the client side, but it is still possible:
{code}
Table tbl = tables.get(name);
if (tbl != null) {
    return CompletableFuture.completedFuture(tbl);
}
{code}
Required to get actual ID from Metastorage before getting a table.
Also, I sure you can to write a test, where Metastorage majority create, drop and create again table with the similar name,  but one Ignite node lagging and has received yet this updates.
Use `WatchListenerInhibitor` to simulate a slow node.

> Issues in implementation of the method TableManager#table(String)
> -----------------------------------------------------------------
>
>                 Key: IGNITE-15718
>                 URL: https://issues.apache.org/jira/browse/IGNITE-15718
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Vladislav Pyatkov
>            Assignee: Mirza Aliev
>            Priority: Major
>              Labels: ignite-3
>             Fix For: 3.0.0-alpha4
>
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> The method (TableManager#table(String)), where we are waiting for a table creation by name, is not correct, because we can see an event about another table with the same name (it is possible when a node is lagging behind of other and is seeing outdated events about drop/create the table, but the name unique only in one time). Need to wait for table creation by ID, because ID is unique for each table even on infinity time.
> Also, after the ticket (IGNITE-15412) where Configuration API was extended of possibility to read a Metasorage value directly, has been fixed, it is time to fix method _TableManager#isTableConfigured(String)_.
> UPD:
> Direct value for _TableManager#isTableConfigured(String)_ has been done in IGNITE-15881



--
This message was sent by Atlassian Jira
(v8.20.1#820001)