ODBC processor default behaviour

[Igor Sapego]

Igniters,

I'm currently working on the ODBC driver. It connects by TCP to the
OdbcProcessor
on the node side. OdbcProcessor is enabled by default and it starts TCP
server on the
specific TCP port or throws exception if the port is busy.

The problem is that such behavior breaks tests that start more than one
node on the
same machine as the second node fails to start because ODBC port is already
taken
by the first node.

Does anyone has ideas what is the best way to fix that?

Best Regards,
Igor

[yzhdanov]

How about outputting warning like and keep retrying in a background thread?

warning - "Failed to bind ODBC processor TCP server to port (retrying every
2 sec) [port=ABC]

--Yakov

2016-02-20 17:38 GMT+03:00 Igor Sapego <[hidden email]>:

> Igniters,
>
> I'm currently working on the ODBC driver. It connects by TCP to the
> OdbcProcessor
> on the node side. OdbcProcessor is enabled by default and it starts TCP
> server on the
> specific TCP port or throws exception if the port is busy.
>
> The problem is that such behavior breaks tests that start more than one
> node on the
> same machine as the second node fails to start because ODBC port is already
> taken
> by the first node.
>
> Does anyone has ideas what is the best way to fix that?
>
> Best Regards,
> Igor
>

[Sergey Kozlov]

What's about to use the port range like TcpDiscoveryVmIpFinder?

On Sat, Feb 20, 2016 at 6:22 PM, Yakov Zhdanov <[hidden email]> wrote:

> How about outputting warning like and keep retrying in a background thread?
>
> warning - "Failed to bind ODBC processor TCP server to port (retrying every
> 2 sec) [port=ABC]
>
> --Yakov
>
> 2016-02-20 17:38 GMT+03:00 Igor Sapego <[hidden email]>:
>
> > Igniters,
> >
> > I'm currently working on the ODBC driver. It connects by TCP to the
> > OdbcProcessor
> > on the node side. OdbcProcessor is enabled by default and it starts TCP
> > server on the
> > specific TCP port or throws exception if the port is busy.
> >
> > The problem is that such behavior breaks tests that start more than one
> > node on the
> > same machine as the second node fails to start because ODBC port is
> already
> > taken
> > by the first node.
> >
> > Does anyone has ideas what is the best way to fix that?
> >
> > Best Regards,
> > Igor
> >
>

--
Sergey Kozlov
GridGain Systems
www.gridgain.com

[yzhdanov]

Not sure if it works. DB connection string should have certain port afaik.

--Yakov

2016-02-20 18:26 GMT+03:00 Sergey Kozlov <[hidden email]>:

> What's about to use the port range like TcpDiscoveryVmIpFinder?
>
> On Sat, Feb 20, 2016 at 6:22 PM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> > How about outputting warning like and keep retrying in a background
> thread?
> >
> > warning - "Failed to bind ODBC processor TCP server to port (retrying
> every
> > 2 sec) [port=ABC]
> >
> > --Yakov
> >
> > 2016-02-20 17:38 GMT+03:00 Igor Sapego <[hidden email]>:
> >
> > > Igniters,
> > >
> > > I'm currently working on the ODBC driver. It connects by TCP to the
> > > OdbcProcessor
> > > on the node side. OdbcProcessor is enabled by default and it starts TCP
> > > server on the
> > > specific TCP port or throws exception if the port is busy.
> > >
> > > The problem is that such behavior breaks tests that start more than one
> > > node on the
> > > same machine as the second node fails to start because ODBC port is
> > already
> > > taken
> > > by the first node.
> > >
> > > Does anyone has ideas what is the best way to fix that?
> > >
> > > Best Regards,
> > > Igor
> > >
> >
>
>
>
> --
> Sergey Kozlov
> GridGain Systems
> www.gridgain.com
>

[Valentin Kulichenko]

I think the behavior of OdbcProcessor should be consistent with
RestProcessor. It tries a configurable range of ports, binds to the first
available and prints it out in the log.

-Val

On Sat, Feb 20, 2016 at 7:33 AM, Yakov Zhdanov <[hidden email]> wrote:

> Not sure if it works. DB connection string should have certain port afaik.
>
> --Yakov
>
> 2016-02-20 18:26 GMT+03:00 Sergey Kozlov <[hidden email]>:
>
> > What's about to use the port range like TcpDiscoveryVmIpFinder?
> >
> > On Sat, Feb 20, 2016 at 6:22 PM, Yakov Zhdanov <[hidden email]>
> > wrote:
> >
> > > How about outputting warning like and keep retrying in a background
> > thread?
> > >
> > > warning - "Failed to bind ODBC processor TCP server to port (retrying
> > every
> > > 2 sec) [port=ABC]
> > >
> > > --Yakov
> > >
> > > 2016-02-20 17:38 GMT+03:00 Igor Sapego <[hidden email]>:
> > >
> > > > Igniters,
> > > >
> > > > I'm currently working on the ODBC driver. It connects by TCP to the
> > > > OdbcProcessor
> > > > on the node side. OdbcProcessor is enabled by default and it starts
> TCP
> > > > server on the
> > > > specific TCP port or throws exception if the port is busy.
> > > >
> > > > The problem is that such behavior breaks tests that start more than
> one
> > > > node on the
> > > > same machine as the second node fails to start because ODBC port is
> > > already
> > > > taken
> > > > by the first node.
> > > >
> > > > Does anyone has ideas what is the best way to fix that?
> > > >
> > > > Best Regards,
> > > > Igor
> > > >
> > >
> >
> >
> >
> > --
> > Sergey Kozlov
> > GridGain Systems
> > www.gridgain.com
> >
>

[Sergi]

I think both approaches are ok if the odbc client understands how should it
deal with connection failures (should it go through a port range or retry a
single port during the period of time).

Sergi

2016-02-23 4:35 GMT+03:00 Valentin Kulichenko <[hidden email]
>:

> I think the behavior of OdbcProcessor should be consistent with
> RestProcessor. It tries a configurable range of ports, binds to the first
> available and prints it out in the log.
>
> -Val
>
> On Sat, Feb 20, 2016 at 7:33 AM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> > Not sure if it works. DB connection string should have certain port
> afaik.
> >
> > --Yakov
> >
> > 2016-02-20 18:26 GMT+03:00 Sergey Kozlov <[hidden email]>:
> >
> > > What's about to use the port range like TcpDiscoveryVmIpFinder?
> > >
> > > On Sat, Feb 20, 2016 at 6:22 PM, Yakov Zhdanov <[hidden email]>
> > > wrote:
> > >
> > > > How about outputting warning like and keep retrying in a background
> > > thread?
> > > >
> > > > warning - "Failed to bind ODBC processor TCP server to port (retrying
> > > every
> > > > 2 sec) [port=ABC]
> > > >
> > > > --Yakov
> > > >
> > > > 2016-02-20 17:38 GMT+03:00 Igor Sapego <[hidden email]>:
> > > >
> > > > > Igniters,
> > > > >
> > > > > I'm currently working on the ODBC driver. It connects by TCP to the
> > > > > OdbcProcessor
> > > > > on the node side. OdbcProcessor is enabled by default and it starts
> > TCP
> > > > > server on the
> > > > > specific TCP port or throws exception if the port is busy.
> > > > >
> > > > > The problem is that such behavior breaks tests that start more than
> > one
> > > > > node on the
> > > > > same machine as the second node fails to start because ODBC port is
> > > > already
> > > > > taken
> > > > > by the first node.
> > > > >
> > > > > Does anyone has ideas what is the best way to fix that?
> > > > >
> > > > > Best Regards,
> > > > > Igor
> > > > >
> > > >
> > >
> > >
> > >
> > > --
> > > Sergey Kozlov
> > > GridGain Systems
> > > www.gridgain.com
> > >
> >
>

[Denis Magda]

Thanks to this old discussion I was finally able to form an understanding on how Ignite ODBC driver connects to the cluster and executes SQL queries over it.

Igor, by some reason we didn’t add any information around ODBC processor and configuration to our documentation page. Let’s fill this gap.
https://issues.apache.org/jira/browse/IGNITE-4184 <https://issues.apache.org/jira/browse/IGNITE-4184>

—
Denis

> On Feb 23, 2016, at 3:04 AM, Sergi Vladykin <[hidden email]> wrote:
>
> I think both approaches are ok if the odbc client understands how should it
> deal with connection failures (should it go through a port range or retry a
> single port during the period of time).
>
> Sergi
>
> 2016-02-23 4:35 GMT+03:00 Valentin Kulichenko <[hidden email]
>> :
>
>> I think the behavior of OdbcProcessor should be consistent with
>> RestProcessor. It tries a configurable range of ports, binds to the first
>> available and prints it out in the log.
>>
>> -Val
>>
>> On Sat, Feb 20, 2016 at 7:33 AM, Yakov Zhdanov <[hidden email]>
>> wrote:
>>
>>> Not sure if it works. DB connection string should have certain port
>> afaik.
>>>
>>> --Yakov
>>>
>>> 2016-02-20 18:26 GMT+03:00 Sergey Kozlov <[hidden email]>:
>>>
>>>> What's about to use the port range like TcpDiscoveryVmIpFinder?
>>>>
>>>> On Sat, Feb 20, 2016 at 6:22 PM, Yakov Zhdanov <[hidden email]>
>>>> wrote:
>>>>
>>>>> How about outputting warning like and keep retrying in a background
>>>> thread?
>>>>>
>>>>> warning - "Failed to bind ODBC processor TCP server to port (retrying
>>>> every
>>>>> 2 sec) [port=ABC]
>>>>>
>>>>> --Yakov
>>>>>
>>>>> 2016-02-20 17:38 GMT+03:00 Igor Sapego <[hidden email]>:
>>>>>
>>>>>> Igniters,
>>>>>>
>>>>>> I'm currently working on the ODBC driver. It connects by TCP to the
>>>>>> OdbcProcessor
>>>>>> on the node side. OdbcProcessor is enabled by default and it starts
>>> TCP
>>>>>> server on the
>>>>>> specific TCP port or throws exception if the port is busy.
>>>>>>
>>>>>> The problem is that such behavior breaks tests that start more than
>>> one
>>>>>> node on the
>>>>>> same machine as the second node fails to start because ODBC port is
>>>>> already
>>>>>> taken
>>>>>> by the first node.
>>>>>>
>>>>>> Does anyone has ideas what is the best way to fix that?
>>>>>>
>>>>>> Best Regards,
>>>>>> Igor
>>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Sergey Kozlov
>>>> GridGain Systems
>>>> www.gridgain.com
>>>>
>>>
>>

[Igor Sapego]

Denis,

Good point. I'll do that.

Best Regards,
Igor

On Tue, Nov 8, 2016 at 1:42 AM, Denis Magda <[hidden email]> wrote:

> Thanks to this old discussion I was finally able to form an understanding
> on how Ignite ODBC driver connects to the cluster and executes SQL queries
> over it.
>
> *Igor*, by some reason we didn’t add any information around ODBC
> processor and configuration to our documentation page. Let’s fill this gap.
> https://issues.apache.org/jira/browse/IGNITE-4184
>
> —
> Denis
>
> On Feb 23, 2016, at 3:04 AM, Sergi Vladykin <[hidden email]>
> wrote:
>
> I think both approaches are ok if the odbc client understands how should it
> deal with connection failures (should it go through a port range or retry a
> single port during the period of time).
>
> Sergi
>
> 2016-02-23 4:35 GMT+03:00 Valentin Kulichenko <
> [hidden email]
>
> :
>
>
> I think the behavior of OdbcProcessor should be consistent with
> RestProcessor. It tries a configurable range of ports, binds to the first
> available and prints it out in the log.
>
> -Val
>
> On Sat, Feb 20, 2016 at 7:33 AM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> Not sure if it works. DB connection string should have certain port
>
> afaik.
>
>
> --Yakov
>
> 2016-02-20 18:26 GMT+03:00 Sergey Kozlov <[hidden email]>:
>
> What's about to use the port range like TcpDiscoveryVmIpFinder?
>
> On Sat, Feb 20, 2016 at 6:22 PM, Yakov Zhdanov <[hidden email]>
> wrote:
>
> How about outputting warning like and keep retrying in a background
>
> thread?
>
>
> warning - "Failed to bind ODBC processor TCP server to port (retrying
>
> every
>
> 2 sec) [port=ABC]
>
> --Yakov
>
> 2016-02-20 17:38 GMT+03:00 Igor Sapego <[hidden email]>:
>
> Igniters,
>
> I'm currently working on the ODBC driver. It connects by TCP to the
> OdbcProcessor
> on the node side. OdbcProcessor is enabled by default and it starts
>
> TCP
>
> server on the
> specific TCP port or throws exception if the port is busy.
>
> The problem is that such behavior breaks tests that start more than
>
> one
>
> node on the
> same machine as the second node fails to start because ODBC port is
>
> already
>
> taken
> by the first node.
>
> Does anyone has ideas what is the best way to fix that?
>
> Best Regards,
> Igor
>
>
>
>
>
> --
> Sergey Kozlov
> GridGain Systems
> www.gridgain.com
>
>
>
>
>