[pgpool-hackers: 1309] Re: [pgpool -C option] need more description?

[Tatsuo Ishii]

> Hi Ishii-san,
> 
> Thanks for prompt response,
> 
>> -----Original Message-----
>> From: Tatsuo Ishii [mailto:ishii at postgresql.org]
>> Sent: Wednesday, January 06, 2016 1:21 PM
>> To: Dangminh Huong(ダンＭ フーン) <huo-dangminh at ys.jp.nec.com>
>> Cc: ishii at postgresql.org; pgpool-hackers at pgpool.net; Yanagisawa
>> Hiroshi(柳澤 博) <hir-yanagisawa at ut.jp.nec.com>
>> Subject: Re: [pgpool-hackers: 1303] Re: [pgpool -C option] need more
>> description?
>> 
>> > Hi Ishii-san,
>> >
>> > Thanks for prompt response,
>> >
>> >> -----Original Message-----
>> >> From: Tatsuo Ishii [mailto:ishii at postgresql.org]
>> >> Sent: Wednesday, January 06, 2016 8:31 AM
>> >> To: Dangminh Huong(ダンＭ フーン) <huo-dangminh at ys.jp.nec.com>;
>> >> Yanagisawa Hiroshi(柳澤 博) <hir-yanagisawa at ut.jp.nec.com>
>> >> Cc: pgpool-hackers at pgpool.net
>> >> Subject: Re: [pgpool-hackers: 1303] Re: [pgpool -C option] need more
>> >> description?
>> >>
>> >> I appeared that the implementation of -c or --clear-cache is quite
>> >> confusing. It does exactly the same thing as -C or --clear-oidmaps. So
>> >> there's no point to have both of options at all.  Moreover -c or
>> >> --clear-cache is not shown when pgpool --help is requested. Apparently
>> >> -c or --clear-cache is useless, and I doubt if we need to have this in
>> >> 3.5.
>> >
>> > I think clear oidmaps should be processed along with clear cache.
>> 
>> No, it's not possible at the moment because pgpool-II cannot
>> distinguish its own cache entries from other entries registered by
>> certain applications on the memcached server. If we could assume that
>> the memcached server is dedicated to pgpool-II, just blowing up all of
>> the cache entries on the memcached server is enough, but that is too
>> simple assumption I guess.
>> 
> 
> Before clear oidmaps, invalidate only cache entries in oid files 
> but not flush out all cache entries in memcached server is also 
> a solution sir?

That's certainly doable.

> It could take a decrease in performance when pgpool start.

Yes, I'm afraid that. Maybe we could limit for a use case where we
have a dedicated memcached server, then we could blow out all the
cache entries at once.

BTW, I want to think about the case which we have 2 or more pgpool
instances and they share same memcached server.  For now, that doesn't
work well because oid files are not shared by those pgpool
instances. Probably oid files should move to the memcached server as
well to solve the problem. This is too late for pgpool-II 3.5 and
could be one of TODO items for 3.6.

> Blowing up all of cache entries by adding -c option also a good idea.
> 
> Best regards,
> ---
> Dang Minh Huong
> NEC Solution Innovators, Ltd.
> http://www.nec-solutioninnovators.co.jp/en/
> 
>> > So, remove -C option and add -c option with clear both of oidmaps
>> > and cache is also a solution in V3.5?
>> >
>> > Best regards,
>> > ---
>> > Dang Minh Huong
>> > NEC Solution Innovators, Ltd.
>> > http://www.nec-solutioninnovators.co.jp/en/
>> >
>> >>
>> >> Best regards,
>> >> --
>> >> Tatsuo Ishii
>> >> SRA OSS, Inc. Japan
>> >> English: http://www.sraoss.co.jp/index_en.php
>> >> Japanese:http://www.sraoss.co.jp
>> >>
>> >> >> Hi all,
>> >> >>
>> >> >> I think below case is depended on the way of using
>> >> >> but is it need to be added more detail at pgpool -C option?
>> >> >> something like "But not clear related cache in memcached server."?
>> >> >>
>> >> >> When memcached is using in "In Memory Cached" mode, pgpool
>> >> >> start with -C option does not clear related cache in cache
>> >> >> server.
>> >> >> Obviously it lead to get the old query result by SELECT
>> >> >> as below.
>> >> >
>> >> > Sounds a fair complaint. Will fix.
>> >> >
>> >> > Best regards,
>> >> > --
>> >> > Tatsuo Ishii
>> >> > SRA OSS, Inc. Japan
>> >> > English: http://www.sraoss.co.jp/index_en.php
>> >> > Japanese:http://www.sraoss.co.jp
>> >> >
>> >> >> ===
>> >> >>
>> >> >> testdb=# select * from testtbl;   -- create cache
>> >> >>  id
>> >> >> ----
>> >> >>   1
>> >> >> (1 row)
>> >> >>
>> >> >> testdb=# \q
>> >> >> ]$ pgpool stop
>> >> >> ]$ pgpool -C -d -n > /var/log/pgpool/log 2>&1 & # start with -C option
>> >> >> ]$ psql -p 9999
>> >> >> psql (9.4.4)
>> >> >> Type "help" for help.
>> >> >>
>> >> >> testdb=# insert into testtbl values(1);
>> >> >> INSERT 0 1
>> >> >> testdb=# select * from testtbl;  -- old query result from cache server
>> >> >>  id
>> >> >> ----
>> >> >>   1
>> >> >> (1 row)
>> >> >>
>> >> >> ===
>> >> >>
>> >> >> Best regards,
>> >> >> ---
>> >> >> Dang Minh Huong
>> >> >> NEC Solution Innovators, Ltd.
>> >> >> http://www.nec-solutioninnovators.co.jp/en/
>> >> >>
>> >> >> _______________________________________________
>> >> >> pgpool-hackers mailing list
>> >> >> pgpool-hackers at pgpool.net
>> >> >> http://www.pgpool.net/mailman/listinfo/pgpool-hackers
>> >> > _______________________________________________
>> >> > pgpool-hackers mailing list
>> >> > pgpool-hackers at pgpool.net
>> >> > http://www.pgpool.net/mailman/listinfo/pgpool-hackers
>> >