[jira] [Commented] (IGNITE-11471) JDBC: Thin driver should start with a random server to connect
[
https://issues.apache.org/jira/browse/IGNITE-11471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16800030#comment-16800030
]
Ivan Pavlukhin commented on IGNITE-11471:
-
[~ibelyakov], [~vozerov], please pay attention that line (from patch) in method
{{JdbcThinTcpIo#nextServerIndex}}
{code}
return (int)(abs(nextIdx) % len);
{code}
can still return negative number because {{Math.abs}} returns
{{Long.MIN_VALUE}} if {{Long.MIN_VALUE}} is passed as argument and {{%}}
operators preserves a sign of a _dividend_ ({{-1 % 2 == -1}}).
> JDBC: Thin driver should start with a random server to connect
> --
>
> Key: IGNITE-11471
> URL: https://issues.apache.org/jira/browse/IGNITE-11471
> Project: Ignite
> Issue Type: Improvement
> Components: jdbc
>Reporter: Stanislav Lukyanov
>Assignee: Igor Belyakov
>Priority: Major
> Labels: newbie
> Fix For: 2.8
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Thin JDBC driver uses a list of server addresses to connect to. Connections
> created in a single JVM will use round-robin to choose the server to connect,
> yielding a uniform distribution. However, the round-robin always starts from
> the first server in the list. It means that multiple JVMs with the same JDBC
> connection URL and one connection each will all connect to the same server.
> Need to choose the first server randomly to have a better distribution.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (IGNITE-11471) JDBC: Thin driver should start with a random server to connect
[ https://issues.apache.org/jira/browse/IGNITE-11471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16797292#comment-16797292 ] Igor Belyakov commented on IGNITE-11471: [~vozerov], In current case RND changing has an effect, because @Before method in the test replaces IgniteUtils.RND instance before the first access to JdbcThinTcpIo. Therefore, JdbcThinTcpIo.IDX_GEN will be initialized by using updated IgniteUtils.RND. Anyway, I agree that it's too tricky solution and exposing IgniteUtils.RND wasn't right. I've reverted last commit. > JDBC: Thin driver should start with a random server to connect > -- > > Key: IGNITE-11471 > URL: https://issues.apache.org/jira/browse/IGNITE-11471 > Project: Ignite > Issue Type: Improvement > Components: jdbc >Reporter: Stanislav Lukyanov >Assignee: Igor Belyakov >Priority: Major > Labels: newbie > Time Spent: 10m > Remaining Estimate: 0h > > Thin JDBC driver uses a list of server addresses to connect to. Connections > created in a single JVM will use round-robin to choose the server to connect, > yielding a uniform distribution. However, the round-robin always starts from > the first server in the list. It means that multiple JVMs with the same JDBC > connection URL and one connection each will all connect to the same server. > Need to choose the first server randomly to have a better distribution. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (IGNITE-11471) JDBC: Thin driver should start with a random server to connect
[ https://issues.apache.org/jira/browse/IGNITE-11471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16796905#comment-16796905 ] Igor Belyakov commented on IGNITE-11471: [~vozerov], random instance for IDX_GEN initialization has been changed to IgniteUtils.RND. Unit test added. > JDBC: Thin driver should start with a random server to connect > -- > > Key: IGNITE-11471 > URL: https://issues.apache.org/jira/browse/IGNITE-11471 > Project: Ignite > Issue Type: Improvement > Components: jdbc >Reporter: Stanislav Lukyanov >Assignee: Igor Belyakov >Priority: Major > Labels: newbie > Time Spent: 10m > Remaining Estimate: 0h > > Thin JDBC driver uses a list of server addresses to connect to. Connections > created in a single JVM will use round-robin to choose the server to connect, > yielding a uniform distribution. However, the round-robin always starts from > the first server in the list. It means that multiple JVMs with the same JDBC > connection URL and one connection each will all connect to the same server. > Need to choose the first server randomly to have a better distribution. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (IGNITE-11471) JDBC: Thin driver should start with a random server to connect
[
https://issues.apache.org/jira/browse/IGNITE-11471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16796895#comment-16796895
]
Ignite TC Bot commented on IGNITE-11471:
{panel:title=-- Run :: All: No blockers
found!|borderStyle=dashed|borderColor=#ccc|titleBGColor=#D6F7C1}{panel}
[TeamCity *-- Run :: All*
Results|https://ci.ignite.apache.org/viewLog.html?buildId=3355627buildTypeId=IgniteTests24Java8_RunAll]
> JDBC: Thin driver should start with a random server to connect
> --
>
> Key: IGNITE-11471
> URL: https://issues.apache.org/jira/browse/IGNITE-11471
> Project: Ignite
> Issue Type: Improvement
> Components: jdbc
>Reporter: Stanislav Lukyanov
>Assignee: Igor Belyakov
>Priority: Major
> Labels: newbie
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Thin JDBC driver uses a list of server addresses to connect to. Connections
> created in a single JVM will use round-robin to choose the server to connect,
> yielding a uniform distribution. However, the round-robin always starts from
> the first server in the list. It means that multiple JVMs with the same JDBC
> connection URL and one connection each will all connect to the same server.
> Need to choose the first server randomly to have a better distribution.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (IGNITE-11471) JDBC: Thin driver should start with a random server to connect
[
https://issues.apache.org/jira/browse/IGNITE-11471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16791518#comment-16791518
]
Ignite TC Bot commented on IGNITE-11471:
{panel:title=-- Run :: All: No blockers
found!|borderStyle=dashed|borderColor=#ccc|titleBGColor=#D6F7C1}{panel}
[TeamCity *-- Run :: All*
Results|https://ci.ignite.apache.org/viewLog.html?buildId=3297736buildTypeId=IgniteTests24Java8_RunAll]
> JDBC: Thin driver should start with a random server to connect
> --
>
> Key: IGNITE-11471
> URL: https://issues.apache.org/jira/browse/IGNITE-11471
> Project: Ignite
> Issue Type: Improvement
> Components: jdbc
>Reporter: Stanislav Lukyanov
>Assignee: Igor Belyakov
>Priority: Major
> Labels: newbie
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Thin JDBC driver uses a list of server addresses to connect to. Connections
> created in a single JVM will use round-robin to choose the server to connect,
> yielding a uniform distribution. However, the round-robin always starts from
> the first server in the list. It means that multiple JVMs with the same JDBC
> connection URL and one connection each will all connect to the same server.
> Need to choose the first server randomly to have a better distribution.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (IGNITE-11471) JDBC: Thin driver should start with a random server to connect
[
https://issues.apache.org/jira/browse/IGNITE-11471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16791475#comment-16791475
]
Ignite TC Bot commented on IGNITE-11471:
{panel:title=-- Run :: All: Possible
Blockers|borderStyle=dashed|borderColor=#ccc|titleBGColor=#F7D6C1}
{color:#d04437}Platform .NET (Long Running){color} [[tests 0 TIMEOUT
|https://ci.ignite.apache.org/viewLog.html?buildId=3297710]]
{panel}
[TeamCity *-- Run :: All*
Results|https://ci.ignite.apache.org/viewLog.html?buildId=3297736buildTypeId=IgniteTests24Java8_RunAll]
> JDBC: Thin driver should start with a random server to connect
> --
>
> Key: IGNITE-11471
> URL: https://issues.apache.org/jira/browse/IGNITE-11471
> Project: Ignite
> Issue Type: Improvement
> Components: jdbc
>Reporter: Stanislav Lukyanov
>Assignee: Igor Belyakov
>Priority: Major
> Labels: newbie
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Thin JDBC driver uses a list of server addresses to connect to. Connections
> created in a single JVM will use round-robin to choose the server to connect,
> yielding a uniform distribution. However, the round-robin always starts from
> the first server in the list. It means that multiple JVMs with the same JDBC
> connection URL and one connection each will all connect to the same server.
> Need to choose the first server randomly to have a better distribution.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (IGNITE-11471) JDBC: Thin driver should start with a random server to connect
[ https://issues.apache.org/jira/browse/IGNITE-11471?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16783581#comment-16783581 ] Stanislav Lukyanov commented on IGNITE-11471: - Choice of the server was random initially, but it was changed to a JVM-local round-robin in IGNITE-9704. Need to use keep the `AtomicLong` for local round-robin, but initialize it with a random value instead of zero. > JDBC: Thin driver should start with a random server to connect > -- > > Key: IGNITE-11471 > URL: https://issues.apache.org/jira/browse/IGNITE-11471 > Project: Ignite > Issue Type: Improvement > Components: jdbc >Reporter: Stanislav Lukyanov >Priority: Major > Labels: newbie > > Thin JDBC driver uses a list of server addresses to connect to. Connections > created in a single JVM will use round-robin to choose the server to connect, > yielding a uniform distribution. However, the round-robin always starts from > the first server in the list. It means that multiple JVMs with the same JDBC > connection URL and one connection each will all connect to the same server. > Need to choose the first server randomly to have a better distribution. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
