ERROR: out of memory DETAIL: Cannot enlarge string buffer containing 1073741822 bytes by 1 more by

[Александр Попов]

Hello. Sorry my English.
I am testing citus. I using next configuration:
1 master,  16 GB memory
4 workers, 16 GB memory,
"PostgreSQL 9.5.1 on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 4.8.4-2ubuntu1~14.04.1) 4.8.4, 64-bit"

On the master, I have I generated the data, using TPCH test. The data has  50 GB.
Each table are distributed on 10 tables.

so, most queries work well.
But, next query has the error:

ERROR:  out of memory
DETAIL:  Cannot enlarge string buffer containing 1073741822 bytes by 1 more bytes.
********** Error **********

ERROR: out of memory
SQL state: 54000
Detail: Cannot enlarge string buffer containing 1073741822 bytes by 1 more bytes.

I changed parameters on master and an workers.

citus.binary_worker_copy_format = true
citus.binary_master_copy_format = true

but, the error  keeps.

The query is:

select
    o_year,
    sum(case
        when nation = 'JAPAN' then volume
        else 0
    end) / sum(volume) as mkt_share
from
    (
        select
            extract(year from o_orderdate) as o_year,
            l_extendedprice * (1 - l_discount) as volume,
            n2.n_name as nation
        from
            part,
            supplier,
            lineitem,
            orders,
            customer,
            nation n1,
            nation n2,
            region
        where
            p_partkey = l_partkey
            and s_suppkey = l_suppkey
            and l_orderkey = o_orderkey
            and o_custkey = c_custkey
            and c_nationkey = n1.n_nationkey
            and n1.n_regionkey = r_regionkey
            and r_name = 'ASIA'
            and s_nationkey = n2.n_nationkey
            and o_orderdate between date '1995-01-01' and date '1996-12-31'
            and p_type = 'PROMO PLATED TIN'
    ) as all_nations
group by
    o_year
order by
    o_year;


What can I  to change to has not this error?


[Metin Doslu]

Hi Александр,

Do you have enough space on your worker nodes? This error is mostly related with malformed input files. For this type of queries, we do lots repartitioning to improve the query performance. Therefore, if you don't have enough space on workers, this would cause to have malformed intermediate files. Could you monitor your disk usage while running this query?

Best,

Metin

--
You received this message because you are subscribed to the Google Groups "citus-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to citus-users...@googlegroups.com.
To post to this group, send email to citus...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/citus-users/6766a034-0e58-4df9-888f-533509102d7e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Александр Попов]

Every worker has more than 500 GB free space.
I am sure, that workers do not use all free space.  This error occurs after 7 seconds of start query.


четверг, 19 мая 2016 г., 13:41:30 UTC+3 пользователь Metin Doslu написал:

[Metin Doslu]

Hi Александр,

I replicated this problem and created an issue on the GitHub to track it. Thank you for reporting this.

https://github.com/citusdata/citus/issues/542

Best,

Metin

To view this discussion on the web visit https://groups.google.com/d/msgid/citus-users/65619aaa-9835-4877-ae66-0eedd8e0fa75%40googlegroups.com.