Re: [Postgres-xc-general] pgxc: snapshot

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

This is related to postgresql.conf parameter max_coordinators and
max_datanodes.    Default value is 16 so you should extend them.
They are specific to XC.

Regards;
----------
Koichi Suzuki

2013/2/19 Arni Sumarlidason <Arn...@md...>:
> The issue seems to be related to the size of the cluster. When attempting to
> initialize 20 nodes. 18, 19, and 20 consistently failed (reproducible). I
> tried a cluster of 17 with fewer errors. I tried 10 with success.
>
>
>
> Best,
>
>
>
> From: Arni Sumarlidason
> Sent: Monday, February 18, 2013 8:14 PM
> To: 'koi...@gm...'; 'Postgres-XC Developers'
> Cc: pos...@li...; mic...@gm...
> Subject: RE: [Postgres-xc-general] pgxc: snapshot
>
>
>
> Mr. Koichi,
>
>
>
> You are right, PGADMIN was the source of these warnings.
>
>
>
> Do you have any idea what would cause the following error:
>
> # Cache lookup failed for node when executing CREATE NODE
>
>
>
> Lastly, I believe there are typos on lines 789, 797 in the pgxc script.
>
>
>
> Best regards,
>
>
>
> -----Original Message-----
> From: koi...@gm... [mailto:koi...@gm...] On Behalf
> Of Koichi Suzuki
> Sent: Monday, February 18, 2013 12:52 AM
> To: Arni Sumarlidason; Postgres-XC Developers
> Subject: Re: [Postgres-xc-general] pgxc: snapshot
>
>
>
> I tried the stress test.  Autovacuum seems to work well.   I also ran
>
> vacuum analyze verbose from psql directly and it worked without
>
> problem during this stress test.   I ran vacuumdb as well.   All
>
> worked without problem.
>
>
>
> I noticed that you used pgAdmin.   Unfortunately, pgAdmin has not been
>
> tuned to work with XC.   I know some of pgAdmin features works well
>
> but others don't.    Could you try to run vacuum from psql or
>
> vacuumdb?    If they don't work, please let me know.
>
>
>
> Best Regards;
>
> ----------
>
> Koichi Suzuki
>
>
>
>
>
> 2013/2/18 Koichi Suzuki <ko...@in...>:
>
>> Nice to hear that pgxc_ctl helps.
>
>>
>
>> As to the warning, I will try to reproduce the problem and fix it.   I
>> need to find a time for it so please forgive me a bit of time.   The test
>> will run many small transactions which will cause autovacuum lauched, as
>> attached.   This test was built as Datanode slave stress test.   I think
>> this may work as autovacuum lauch test.   I will test it with four
>> coordinators and four datanodes, and four gtm_proxies as well.   Whole test
>> will take about a couple of hours with five of six-core Xeon servers (one
>> for GTM).
>
>>
>
>> Do you think this makes sense to reproduce your problem?
>
>>
>
>> I will run it both on master and REL1_0_STABLE.
>
>>
>
>> Regards;
>
>> ---
>
>> Koichi
>
>>
>
>> On Sat, 16 Feb 2013 19:32:11 +0000
>
>> Arni Sumarlidason <Arn...@md...> wrote:
>
>>
>
>>> Koichi, and others,
>
>>>
>
>>> I spun some fresh VMs and ran your script with the identical outcome, GTM
>>> Snapshot warnings from the auto vacuum launcher.
>
>>> Please advise.
>
>>>
>
>>>
>
>>> Thank you for your script, it does make life easier!!
>
>>>
>
>>> Best,
>
>>>
>
>>> -----Original Message-----
>
>>> From: Koichi Suzuki [mailto:ko...@in...]
>
>>> Sent: Friday, February 15, 2013 4:11 AM
>
>>> To: Arni Sumarlidason
>
>>> Cc: Michael Paquier; koi...@gm...;
>
>>> pos...@li...
>
>>> Subject: Re: [Postgres-xc-general] pgxc: snapshot
>
>>>
>
>>> If you're not sure about the configuration, please try pgxc_ctl
>
>>> available at
>
>>>
>
>>> git://github.com/koichi-szk/PGXC-Tools.git
>
>>>
>
>>> This is bash script (I'm rewriting into C now) so it will help to
>>> understand how to configure XC.
>
>>>
>
>>> Regards;
>
>>> ---
>
>>> Koichi Suzuki
>
>>>
>
>>> On Fri, 15 Feb 2013 04:22:49 +0000
>
>>> Arni Sumarlidason <Arn...@md...> wrote:
>
>>>
>
>>> > Thank you both for fast response!!
>
>>> >
>
>>> > RE: Koichi Suzuki
>
>>> > I downloaded the git this afternoon.
>
>>> >
>
>>> > RE: Michael Paquier
>
>>> >
>
>>> > -          Confirm it is from the datanode's log.
>
>>> >
>
>>> > -          Both coord & datanode connect via the same gtm_proxy on
>>> > localhost
>
>>> >
>
>>> > These are my simplified configs, the only change I make on each
>
>>> > node is the nodename, PG_HBA
>
>>> > local   all             all
>>> > trust
>
>>> > host    all             all             127.0.0.1/32
>>> > trust
>
>>> > host    all             all             ::1/128
>>> > trust
>
>>> > host    all             all             10.100.170.0/24
>>> > trust
>
>>> >
>
>>> > COORD
>
>>> > pgxc_node_name = 'coord01'
>
>>> > listen_addresses = '*'
>
>>> > port = 5432
>
>>> > max_connections = 200
>
>>> >
>
>>> > gtm_port = 6666
>
>>> > gtm_host = 'localhost'
>
>>> > pooler_port = 6670
>
>>> >
>
>>> > shared_buffers = 32MB
>
>>> > work_mem = 1MB
>
>>> > maintenance_work_mem = 16MB
>
>>> > max_stack_depth = 2MB
>
>>> >
>
>>> > log_timezone = 'US/Eastern'
>
>>> > datestyle = 'iso, mdy'
>
>>> > timezone = 'US/Eastern'
>
>>> > lc_messages = 'en_US.UTF-8'
>
>>> > lc_monetary = 'en_US.UTF-8'
>
>>> > lc_numeric = 'en_US.UTF-8'
>
>>> > lc_time = 'en_US.UTF-8'
>
>>> > default_text_search_config = 'pg_catalog.english'
>
>>> >
>
>>> > DATA
>
>>> > pgxc_node_name = 'data01'
>
>>> > listen_addresses = '*'
>
>>> > port = 5433
>
>>> > max_connections = 200
>
>>> >
>
>>> > gtm_port = 6666
>
>>> > gtm_host = 'localhost'
>
>>> >
>
>>> > shared_buffers = 32MB
>
>>> > work_mem = 1MB
>
>>> > maintenance_work_mem = 16MB
>
>>> > max_stack_depth = 2MB
>
>>> >
>
>>> > log_timezone = 'US/Eastern'
>
>>> > datestyle = 'iso, mdy'
>
>>> > timezone = 'US/Eastern'
>
>>> > lc_messages = 'en_US.UTF-8'
>
>>> > lc_monetary = 'en_US.UTF-8'
>
>>> > lc_numeric = 'en_US.UTF-8'
>
>>> > lc_time = 'en_US.UTF-8'
>
>>> > default_text_search_config = 'pg_catalog.english'
>
>>> >
>
>>> > PROXY
>
>>> > Nodename = 'proxy01'
>
>>> > listen_addresses = '*'
>
>>> > port = 6666
>
>>> > gtm_host = '10.100.170.10'
>
>>> > gtm_port = 6666
>
>>> >
>
>>> >
>
>>> > best,
>
>>> >
>
>>> > Arni
>
>>> >
>
>>> > From: Michael Paquier [mailto:mic...@gm...]
>
>>> > Sent: Thursday, February 14, 2013 11:06 PM
>
>>> > To: Arni Sumarlidason
>
>>> > Cc: pos...@li...
>
>>> > Subject: Re: [Postgres-xc-general] pgxc: snapshot
>
>>> >
>
>>> >
>
>>> > On Fri, Feb 15, 2013 at 12:57 PM, Arni Sumarlidason
>>> > <Arn...@md...<mailto:Arn...@md...>> wrote:
>
>>> > Hi Everyone!
>
>>> >
>
>>> > I am getting these errors, "Warning: do not have a gtm snapshot
>>> > available"[1]. After researching I found posts about the auto vacuum causing
>>> > these errors, is this fix or work in progress? Also, I am seeing them
>>> > without the CONTEXT: automatic vacuum message too. Is this something to
>>> > worry about? Cluster seems to be functioning normally.
>
>>> >
>
>>> > Vacuum and analyze from pgadmin looks like this,
>
>>> > INFO:  vacuuming "public.table"
>
>>> > INFO:  "table": found 0 removable, 0 nonremovable row versions in 0
>
>>> > pages
>
>>> > DETAIL:  0 dead row versions cannot be removed yet.
>
>>> > CPU 0.00s/0.00u sec elapsed 0.00 sec.
>
>>> > INFO:  analyzing "public.table"
>
>>> > INFO:  "table": scanned 0 of 0 pages, containing 0 live rows and 0
>
>>> > dead rows; 0 rows in sample, 0 estimated total rows Total query
>>> > runtime: 15273 ms.
>
>>> >
>
>>> > Should we use execute direct to perform maintenance?
>
>>> > No. Isn't this happening on a Datanode?
>
>>> > Be sure first to set gtm_host and gtm_port in postgresql.conf of all
>>> > the nodes, Coordinator and Datanode included. GXID and snapshots are fetched
>>> > of course on Coordinator for normal transaction run but also on all the
>>> > nodes for autovacuum.
>
>>> > --
>
>>> > Michael
>
>>>