Hello,
I've asked this question in serusers and serdevel. Noone answered, so I thought I'd ask it here about openser. I haven't tried openser yet, it's on my todo list. I saw that openser implemented connection pool for mysql backend. What about postgres? My guess is that every module using databse opens a separate connection and the number of concurrent connections will be the number of such modules multiplied by number of children? Is it true? If so, are there any plans for connection pool for postgres? Also it's been noticed and confirmed by Andrei Pelinescu-Onciul that there are problems linking ser to multi-threaded libraries like libpq. Has it been addressed by openser? Also I'm curious whether number of children in openser is limited to a given startup value or it is able to fork additional children as needed? Thank you,
Michael
Hello,
On 07/21/05 23:20, Michael Ulitskiy wrote:
Hello,
I've asked this question in serusers and serdevel. Noone answered, so I thought I'd ask it here about openser. I haven't tried openser yet, it's on my todo list. I saw that openser implemented connection pool for mysql backend. What about postgres?
The maintainer of this module has not implemented the connection pool yet. It was no activity for quite some time of postgres module,but now seems that many users started using it, so we have put it in the roadmap to review. Unfortunately I have no experience with postgres, so the priority is quite low in my list.
My guess is that every module using databse opens a separate connection and the number of concurrent connections will be the number of such modules multiplied by number of children? Is it true?
yes.
If so, are there any plans for connection pool for postgres?
see above.
Also it's been noticed and confirmed by Andrei Pelinescu-Onciul that there are problems linking ser to multi-threaded libraries like libpq. Has it been addressed by openser?
is the same problem as in ser. It is the same module, we just fixed some warnings and a few bugs reported by users.
Also I'm curious whether number of children in openser is limited to a given startup value or it is able to fork additional children as needed?
It is limited to startup value, it does not fork additional children if needed. The maximum number of children is 32. By default it is 8.
Daniel
Thank you,
Michael
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Daniel, Thanks for the reply. Please see inline.
On Friday 22 July 2005 04:31 am, Daniel-Constantin Mierla wrote:
Hello,
On 07/21/05 23:20, Michael Ulitskiy wrote:
Hello,
I've asked this question in serusers and serdevel. Noone answered, so I thought I'd ask it here about openser. I haven't tried openser yet, it's on my todo list. I saw that openser implemented connection pool for mysql backend. What about postgres?
The maintainer of this module has not implemented the connection pool yet. It was no activity for quite some time of postgres module,but now seems that many users started using it, so we have put it in the roadmap to review. Unfortunately I have no experience with postgres, so the priority is quite low in my list.
I see. That's a pity. BTW I can't find it in the roadmap available online.
My guess is that every module using databse opens a separate connection and the number of concurrent connections will be the number of such modules multiplied by number of children? Is it true?
yes.
If so, are there any plans for connection pool for postgres?
see above.
Also it's been noticed and confirmed by Andrei Pelinescu-Onciul that there are problems linking ser to multi-threaded libraries like libpq. Has it been addressed by openser?
is the same problem as in ser. It is the same module, we just fixed some warnings and a few bugs reported by users.
Then I just wanted to let you know that changes suggested by Dmitry Semyonov at http://mail.iptel.org/pipermail/serdev/2005-July/005148.html worked for me. At least it helped with "log_stderr=no" problem.
Also I'm curious whether number of children in openser is limited to a given startup value or it is able to fork additional children as needed?
It is limited to startup value, it does not fork additional children if needed. The maximum number of children is 32. By default it is 8.
Hm. I definitely remember reading on mailing list suggestions to run ser with as many as hundreds children to avoid situation when all workers are busy processing requests. That's not a valid suggestion? What would happen to messages when there're no idle children? Will they be dropped? Please don't get me wrong. I'm not to undervalue your work. I just want to understand the limitation of the software I'm considering to use.
Daniel
Thank you,
Michael
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
On 07/22/05 19:50, Michael Ulitskiy wrote:
Daniel, Thanks for the reply. Please see inline.
On Friday 22 July 2005 04:31 am, Daniel-Constantin Mierla wrote:
Hello,
On 07/21/05 23:20, Michael Ulitskiy wrote:
Hello,
I've asked this question in serusers and serdevel. Noone answered, so I thought I'd ask it here about openser. I haven't tried openser yet, it's on my todo list. I saw that openser implemented connection pool for mysql backend. What about postgres?
The maintainer of this module has not implemented the connection pool yet. It was no activity for quite some time of postgres module,but now seems that many users started using it, so we have put it in the roadmap to review. Unfortunately I have no experience with postgres, so the priority is quite low in my list.
I see. That's a pity. BTW I can't find it in the roadmap available online.
The review explicitly mentions the migration to the memory manager used by ser.
My guess is that every module using databse opens a separate connection and the number of concurrent connections will be the number of such modules multiplied by number of children? Is it true?
yes.
If so, are there any plans for connection pool for postgres?
see above.
Also it's been noticed and confirmed by Andrei Pelinescu-Onciul that there are problems linking ser to multi-threaded libraries like libpq. Has it been addressed by openser?
is the same problem as in ser. It is the same module, we just fixed some warnings and a few bugs reported by users.
Then I just wanted to let you know that changes suggested by Dmitry Semyonov at http://mail.iptel.org/pipermail/serdev/2005-July/005148.html worked for me. At least it helped with "log_stderr=no" problem.
ok, good to know. Maybe an entry in the FAQ will help the others in the future.
Also I'm curious whether number of children in openser is limited to a given startup value or it is able to fork additional children as needed?
It is limited to startup value, it does not fork additional children if needed. The maximum number of children is 32. By default it is 8.
Hm. I definitely remember reading on mailing list suggestions to run ser with as many as hundreds children to avoid situation when all workers are busy processing requests. That's not a valid suggestion? What would happen to messages when there're no idle children? Will they be dropped?
sorry, that was a misleading info from my side, I was thinking to a wrong bitmask present in openser. You can start as many as children you want.
Daniel
Please don't get me wrong. I'm not to undervalue your work. I just want to understand the limitation of the software I'm considering to use.
Daniel
Thank you,
Michael
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
On Friday 22 July 2005 01:06 pm, Daniel-Constantin Mierla wrote:
Hello,
I've asked this question in serusers and serdevel. Noone answered, so I thought I'd ask it here about openser. I haven't tried openser yet, it's on my todo list. I saw that openser implemented connection pool for mysql backend. What about postgres?
The maintainer of this module has not implemented the connection pool yet. It was no activity for quite some time of postgres module,but now seems that many users started using it, so we have put it in the roadmap to review. Unfortunately I have no experience with postgres, so the priority is quite low in my list.
I see. That's a pity. BTW I can't find it in the roadmap available online.
The review explicitly mentions the migration to the memory manager used by ser.
Sorry, it wasn't clear for me that this is about connection pool implementation.
sorry, that was a misleading info from my side, I was thinking to a wrong bitmask present in openser. You can start as many as children you want.
Thanks for the info.
Daniel
On 07/22/05 20:54, Michael Ulitskiy wrote:
On Friday 22 July 2005 01:06 pm, Daniel-Constantin Mierla wrote:
Hello,
I've asked this question in serusers and serdevel. Noone answered, so I thought I'd ask it here about openser. I haven't tried openser yet, it's on my todo list. I saw that openser implemented connection pool for mysql backend. What about postgres?
The maintainer of this module has not implemented the connection pool yet. It was no activity for quite some time of postgres module,but now seems that many users started using it, so we have put it in the roadmap to review. Unfortunately I have no experience with postgres, so the priority is quite low in my list.
I see. That's a pity. BTW I can't find it in the roadmap available online.
The review explicitly mentions the migration to the memory manager used by ser.
Sorry, it wasn't clear for me that this is about connection pool implementation.
because I have no much knowledge about this module -- the stuff about the memory manager I discovered when Ingo Wolfsberger told us about a lot of warnings issued by this module. But when it is going to be reviewed, it should be aligned to latest DB API (implemented now by mysql module).
Daniel
sorry, that was a misleading info from my side, I was thinking to a wrong bitmask present in openser. You can start as many as children you want.
Thanks for the info.
Daniel
-
Daniel-Constantin Mierla -
Michael Ulitskiy