You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@trafodion.apache.org by Dave Birdsall <da...@esgyn.com> on 2017/01/04 21:49:36 UTC
Debugging issues in mxosrvr
Hi,
I have an executor issue that occurs only in mxosrvr, not in sqlci. So I want to debug it in mxosrvr, using gdb.
I gather that DCS has several timeouts that will cause mxosrvr to be killed or to kill itself if they go too long. I'm wondering which of these I need to change in order to have a stable gdb session. I see the following listed in the DCS manual (here: http://trafodion.apache.org/docs/dcs_reference/index.html):
dcs.master.listener.request.timeout
Description
Listener Request timeout. Default 30 seconds.
Default
30000
dcs.master.listener.selector.timeout
Description
Listener Selector timeout. Default 10 seconds.
Default
10000
dcs.server.user.program.zookeeper.session.timeout
Description
User program ZooKeeper session timeout. Default 180 seconds.
Default
180
zookeeper.session.timeout
Description
ZooKeeper session timeout. dcs passes this to the zk quorum as suggested maximum time for a session (This setting becomes zookeeper's 'maxSessionTimeout'). See http://hadoop.apache.org/zookeeper/docs/current/zookeeperProgrammers.html#ch_zkSessions "The client sends a requested timeout, the server responds with the timeout that it can give the client. " In milliseconds.
Default
180000
Also, are there any other settings I should set to get a good debugging experience? (I'm already aware of turning off Repository writes by setting dcs.server.user.program.statistics.enabled to 'false'.)
Thanks,
Dave
Re: Debugging issues in mxosrvr
Posted by Rinka Singh <ri...@gmail.com>.
Thanks.
I'll review it as a "novice" bug fixer. Given that I am still trying to
understand Trafordion.
On Friday 06 January 2017 03:20 AM, Dave Birdsall wrote:
> Hi Rinka,
>
> Good suggestion! I've added this and some other mxosrvr debugging tips to the following Trafodion wiki page: https://cwiki.apache.org/confluence/display/TRAFODION/Debugging+Tips.
>
> Hi Community,
>
> Those of you who are knowledgeable about mxosrvr debugging: Please take a look at this page and correct any errors I made. And add any other tips you think are worth while.
>
> Dave
>
> -----Original Message-----
> From: Rinka Singh [mailto:rinka.singh@gmail.com]
> Sent: Wednesday, January 4, 2017 9:16 PM
> To: dev@trafodion.incubator.apache.org
> Subject: Re: Debugging issues in mxosrvr
>
> This might be good information to add to the documentation - checked, its not there. Also a brief writeup on debugging (basically the settings and things to consider).
>
> Thanks,
> On Thursday 05 January 2017 03:49 AM, Anuradha Hegde wrote:
>> To debug mxosrvr you can set zktimeout <property>
>> <name>dcs.server.user.program.zookeeper.session.timeout</name>
>> <value>3600</value>
>> </property>
>> <property>
>> <name>zookeeper.session.timeout</name>
>> <value>3600000</value>
>> </property>
>>
>> Anu
>>
>> -----Original Message-----
>> From: Dave Birdsall [mailto:dave.birdsall@esgyn.com]
>> Sent: Wednesday, January 4, 2017 1:50 PM
>> To: dev@trafodion.incubator.apache.org
>> Subject: Debugging issues in mxosrvr
>>
>> Hi,
>>
>> I have an executor issue that occurs only in mxosrvr, not in sqlci. So I want to debug it in mxosrvr, using gdb.
>>
>> I gather that DCS has several timeouts that will cause mxosrvr to be killed or to kill itself if they go too long. I'm wondering which of these I need to change in order to have a stable gdb session. I see the following listed in the DCS manual (here: http://trafodion.apache.org/docs/dcs_reference/index.html):
>>
>> dcs.master.listener.request.timeout
>> Description
>> Listener Request timeout. Default 30 seconds.
>> Default
>> 30000
>> dcs.master.listener.selector.timeout
>> Description
>> Listener Selector timeout. Default 10 seconds.
>> Default
>> 10000
>> dcs.server.user.program.zookeeper.session.timeout
>> Description
>> User program ZooKeeper session timeout. Default 180 seconds.
>> Default
>> 180
>> zookeeper.session.timeout
>> Description
>>
>> ZooKeeper session timeout. dcs passes this to the zk quorum as suggested maximum time for a session (This setting becomes zookeeper's 'maxSessionTimeout'). See http://hadoop.apache.org/zookeeper/docs/current/zookeeperProgrammers.html#ch_zkSessions "The client sends a requested timeout, the server responds with the timeout that it can give the client. " In milliseconds.
>> Default
>>
>> 180000
>> Also, are there any other settings I should set to get a good
>> debugging experience? (I'm already aware of turning off Repository
>> writes by setting dcs.server.user.program.statistics.enabled to
>> 'false'.)
>>
>> Thanks,
>>
>> Dave
>>
RE: Debugging issues in mxosrvr
Posted by Dave Birdsall <da...@esgyn.com>.
Hi Rinka,
Good suggestion! I've added this and some other mxosrvr debugging tips to the following Trafodion wiki page: https://cwiki.apache.org/confluence/display/TRAFODION/Debugging+Tips.
Hi Community,
Those of you who are knowledgeable about mxosrvr debugging: Please take a look at this page and correct any errors I made. And add any other tips you think are worth while.
Dave
-----Original Message-----
From: Rinka Singh [mailto:rinka.singh@gmail.com]
Sent: Wednesday, January 4, 2017 9:16 PM
To: dev@trafodion.incubator.apache.org
Subject: Re: Debugging issues in mxosrvr
This might be good information to add to the documentation - checked, its not there. Also a brief writeup on debugging (basically the settings and things to consider).
Thanks,
On Thursday 05 January 2017 03:49 AM, Anuradha Hegde wrote:
> To debug mxosrvr you can set zktimeout <property>
> <name>dcs.server.user.program.zookeeper.session.timeout</name>
> <value>3600</value>
> </property>
> <property>
> <name>zookeeper.session.timeout</name>
> <value>3600000</value>
> </property>
>
> Anu
>
> -----Original Message-----
> From: Dave Birdsall [mailto:dave.birdsall@esgyn.com]
> Sent: Wednesday, January 4, 2017 1:50 PM
> To: dev@trafodion.incubator.apache.org
> Subject: Debugging issues in mxosrvr
>
> Hi,
>
> I have an executor issue that occurs only in mxosrvr, not in sqlci. So I want to debug it in mxosrvr, using gdb.
>
> I gather that DCS has several timeouts that will cause mxosrvr to be killed or to kill itself if they go too long. I'm wondering which of these I need to change in order to have a stable gdb session. I see the following listed in the DCS manual (here: http://trafodion.apache.org/docs/dcs_reference/index.html):
>
> dcs.master.listener.request.timeout
> Description
> Listener Request timeout. Default 30 seconds.
> Default
> 30000
> dcs.master.listener.selector.timeout
> Description
> Listener Selector timeout. Default 10 seconds.
> Default
> 10000
> dcs.server.user.program.zookeeper.session.timeout
> Description
> User program ZooKeeper session timeout. Default 180 seconds.
> Default
> 180
> zookeeper.session.timeout
> Description
>
> ZooKeeper session timeout. dcs passes this to the zk quorum as suggested maximum time for a session (This setting becomes zookeeper's 'maxSessionTimeout'). See http://hadoop.apache.org/zookeeper/docs/current/zookeeperProgrammers.html#ch_zkSessions "The client sends a requested timeout, the server responds with the timeout that it can give the client. " In milliseconds.
> Default
>
> 180000
> Also, are there any other settings I should set to get a good
> debugging experience? (I'm already aware of turning off Repository
> writes by setting dcs.server.user.program.statistics.enabled to
> 'false'.)
>
> Thanks,
>
> Dave
>
Re: Debugging issues in mxosrvr
Posted by Rinka Singh <ri...@gmail.com>.
This might be good information to add to the documentation - checked,
its not there. Also a brief writeup on debugging (basically the
settings and things to consider).
Thanks,
On Thursday 05 January 2017 03:49 AM, Anuradha Hegde wrote:
> To debug mxosrvr you can set zktimeout
> <property>
> <name>dcs.server.user.program.zookeeper.session.timeout</name>
> <value>3600</value>
> </property>
> <property>
> <name>zookeeper.session.timeout</name>
> <value>3600000</value>
> </property>
>
> Anu
>
> -----Original Message-----
> From: Dave Birdsall [mailto:dave.birdsall@esgyn.com]
> Sent: Wednesday, January 4, 2017 1:50 PM
> To: dev@trafodion.incubator.apache.org
> Subject: Debugging issues in mxosrvr
>
> Hi,
>
> I have an executor issue that occurs only in mxosrvr, not in sqlci. So I want to debug it in mxosrvr, using gdb.
>
> I gather that DCS has several timeouts that will cause mxosrvr to be killed or to kill itself if they go too long. I'm wondering which of these I need to change in order to have a stable gdb session. I see the following listed in the DCS manual (here: http://trafodion.apache.org/docs/dcs_reference/index.html):
>
> dcs.master.listener.request.timeout
> Description
> Listener Request timeout. Default 30 seconds.
> Default
> 30000
> dcs.master.listener.selector.timeout
> Description
> Listener Selector timeout. Default 10 seconds.
> Default
> 10000
> dcs.server.user.program.zookeeper.session.timeout
> Description
> User program ZooKeeper session timeout. Default 180 seconds.
> Default
> 180
> zookeeper.session.timeout
> Description
>
> ZooKeeper session timeout. dcs passes this to the zk quorum as suggested maximum time for a session (This setting becomes zookeeper's 'maxSessionTimeout'). See http://hadoop.apache.org/zookeeper/docs/current/zookeeperProgrammers.html#ch_zkSessions "The client sends a requested timeout, the server responds with the timeout that it can give the client. " In milliseconds.
> Default
>
> 180000
> Also, are there any other settings I should set to get a good debugging experience? (I'm already aware of turning off Repository writes by setting dcs.server.user.program.statistics.enabled to 'false'.)
>
> Thanks,
>
> Dave
>
RE: Debugging issues in mxosrvr
Posted by Anuradha Hegde <an...@esgyn.com>.
To debug mxosrvr you can set zktimeout
<property>
<name>dcs.server.user.program.zookeeper.session.timeout</name>
<value>3600</value>
</property>
<property>
<name>zookeeper.session.timeout</name>
<value>3600000</value>
</property>
Anu
-----Original Message-----
From: Dave Birdsall [mailto:dave.birdsall@esgyn.com]
Sent: Wednesday, January 4, 2017 1:50 PM
To: dev@trafodion.incubator.apache.org
Subject: Debugging issues in mxosrvr
Hi,
I have an executor issue that occurs only in mxosrvr, not in sqlci. So I want to debug it in mxosrvr, using gdb.
I gather that DCS has several timeouts that will cause mxosrvr to be killed or to kill itself if they go too long. I'm wondering which of these I need to change in order to have a stable gdb session. I see the following listed in the DCS manual (here: http://trafodion.apache.org/docs/dcs_reference/index.html):
dcs.master.listener.request.timeout
Description
Listener Request timeout. Default 30 seconds.
Default
30000
dcs.master.listener.selector.timeout
Description
Listener Selector timeout. Default 10 seconds.
Default
10000
dcs.server.user.program.zookeeper.session.timeout
Description
User program ZooKeeper session timeout. Default 180 seconds.
Default
180
zookeeper.session.timeout
Description
ZooKeeper session timeout. dcs passes this to the zk quorum as suggested maximum time for a session (This setting becomes zookeeper's 'maxSessionTimeout'). See http://hadoop.apache.org/zookeeper/docs/current/zookeeperProgrammers.html#ch_zkSessions "The client sends a requested timeout, the server responds with the timeout that it can give the client. " In milliseconds.
Default
180000
Also, are there any other settings I should set to get a good debugging experience? (I'm already aware of turning off Repository writes by setting dcs.server.user.program.statistics.enabled to 'false'.)
Thanks,
Dave