postgres-xc-general Mailing List for Postgres-XC

This looks strange.

We run DBT-1 with multiple coordinators, where we open more than 10
connections, and don't see this phenomenon. Is there some load balancer or
some tool sitting between coordinators and application?


On Wed, Apr 10, 2013 at 1:32 AM, Arni Sumarlidason <
Arn...@md...> wrote:

>  Mr. Bapat,****
>
> ** **
>
> Sorry for the slow response, I have been sick.****
>
> ** **
>
> I can open five connections to a coordinator, upon the sixth connection
> all connections to any coordinator are dropped. I thought this might have
> to do with the connection max specified in postgresql.conf, I adjusted it,
> same results. No actions are necessary to recover from the crash, just open
> a new connection.****
>
> ** **
>
> *From:* Ashutosh Bapat [mailto:ash...@en...]
> *Sent:* Saturday, April 06, 2013 12:19 AM
>
> *To:* Arni Sumarlidason
> *Cc:* pos...@li...
> *Subject:* Re: [Postgres-xc-general] Heavy update utilization****
>
> ** **
>
> Can you please elaborate more on what kind of cascade failure you are
> seeing? What are the symptoms? Is it crash? How bad it is? Does it affect
> only the running transaction or entire systems needing reboot or needs data
> clusters to be rebuilt?****
>
> ** **
>
> On Fri, Apr 5, 2013 at 7:12 PM, Arni Sumarlidason <
> Arn...@md...> wrote:****
>
> Does that also explain the cascade failure?****
>
>  ****
>
> Is there a cursor limit per coordinator?****
>
>  ****
>
> *From:* Ashutosh Bapat [mailto:ash...@en...]
> *Sent:* Friday, April 05, 2013 1:37 AM
> *To:* Arni Sumarlidason
> *Cc:* pos...@li...
> *Subject:* Re: [Postgres-xc-general] Heavy update utilization****
>
>  ****
>
> Inheritance is a problem here. Otherwise, we try to send DMLs involving a
> single table directly to the datanode, without any planning at coordinator.
> Unfortunately in this case since inheritance is involved, is row is
> modified first on coordinator and then changed at the datanode.****
>
> If you are going to use inheritance a lot, may be you want to take up
> optimizations for inheritance cases (this is in the context of other
> thread, where you have asked for a todo list.)****
>
>  ****
>
> On Fri, Apr 5, 2013 at 10:18 AM, Arni Sumarlidason <
> Arn...@md...> wrote:****
>
> Schema same as before, we left our schema inherited because we don't want
> to reload our data when support is pushed out. ****
>
>  ****
>
> Think inheritance is also causing this issue as well?****
>
>  ****
>
> Statement: Update table set col = "stuff" where Id = 447938****
>
>  ****
>
> Also might be worth mentioning we are using psycopg2 python connector. ***
> *
>
>  ****
>
> Thank you for your advice.
>
> Sent from my iPhone****
>
>
> On Apr 4, 2013, at 11:57 PM, "Ashutosh Bapat" <
> ash...@en...> wrote:****
>
>  Can you please post your update query here with all the relevant
> definitions like table schema etc.?****
>
>  ****
>
> On Fri, Apr 5, 2013 at 3:15 AM, Arni Sumarlidason <
> Arn...@md...> wrote:****
>
> General,****
>
>  ****
>
> We had a problem doing updates on our database today, it seems like
> updates are a pretty heavy operation for the coordinator. They tack our
> machines at 60-70% utilization[1] with 5 update cursors, and for reasons
> unknown a 6th connection causes a cascade failure across all
> coordinators. ****
>
>  ****
>
>  ****
>
> Do you know what would cause this?****
>
>  ****
>
> [1]****
>
> http://www.sumarlidason.com/perm/130404/load8.PNG<https://console.mxlogic.com/redir/?zC763hOYqejhOrv7cIcECzAQsCM0l6Rtyv93qlxO-bv5n4mrInoM-KEfs5wo4zt-psLuZXTLsTsTJ2rsWQ9YgRrpwHUj-aQpkZ0xOVIs-eodEIKfEII9ILCTXCMnWhEwdbomHip6fQbxZyV7Uv7jPiWq80L-CiNEwrmgSvCy1KszV2CPBm1KvxYY1NJ4SOrpop73zhO-Ur3Fk-amvt>- Four machines at bottom of screen are coordinators, which correspond to
> the thick utilization in the graph.****
>
>  ****
>
>
>
> ------------------------------------------------------------------------------
> Minimize network downtime and maximize team effectiveness.
> Reduce network management and security costs.Learn how to hire
> the most talented Cisco Certified professionals. Visit the
> Employer Resources Portal
> http://www.cisco.com/web/learning/employer_resources/index.html<https://console.mxlogic.com/redir/?zC763hOYqejhOrv7cIcECzAQsCM0lFkJnBY05x_X64E4sZlo-FrUyIvFlKdLFw5Y3zt-psLuZXTLsTsTJ2rsWQ9YgRrpwHUj-aQpkZ0xOVIs-eodEIKfEII9ILCTXCMnWhEwdbomHip6fQbxZyV7Uv7jPiWq80L-CiNEwrmgSvCy1KszV2CPBm1KvxYY1NJcSOrpop73zhO-UrlC3O>
> _______________________________________________
> Postgres-xc-general mailing list
> Pos...@li...
> https://lists.sourceforge.net/lists/listinfo/postgres-xc-general<https://console.mxlogic.com/redir/?2eosod7bNEVd79JYsOMOyqejhOr9PCJhbcfBiteFlKdLt00_MdM-l9QM-l9OwXn5GQChzZ2UvoKh-7NQYTvCnbTLuZXTdTdXgCTeJ2v4dmSoa-4_yJ6lfg8sKr7fzC3qbbzWbb2rbVJ-VI5-Aq83iS5GQChzZ2UvoKh-7NQYQKCy0b_FAIq86RAdDVEwrD8-gFIVlwrDUvf0srpdICSm6hMUQsLK6XlrO>
> ****
>
>
>
>
> --
> Best Wishes,
> Ashutosh Bapat
> EntepriseDB Corporation
> The Enterprise Postgres Company****
>
>
>
>
> --
> Best Wishes,
> Ashutosh Bapat
> EntepriseDB Corporation
> The Enterprise Postgres Company****
>
>
>
>
> --
> Best Wishes,
> Ashutosh Bapat
> EntepriseDB Corporation
> The Enterprise Postgres Company****
>



-- 
Best Wishes,
Ashutosh Bapat
EntepriseDB Corporation
The Enterprise Postgres Company