Hi Guys,

I am running following Pig script in Pig 0.8 version

page_events = LOAD '/user/sgehlot/day=2011-05-10' as
(event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_number:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:int,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbound_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_id:int,ontology_node_id:int,content_namespace_id:int,product_namespace_id:int,transparent_edition_id:int,default_edition_id:int,event_seq_num:int,is_last_page:int,is_new_user:int,page_duration:int,page_seq_num:int,session_id:long,time_since_sess_start:int,reg_cookie:chararray,urs_app_id:int,is_reg_user:int,edition_id:int,user_agent_id:int,page_type_key:int,referrer_id:int,channel_id:int,level2_id:int,level3_id:int,brand_id:int,content_key:int,product_key:int,edition_key:int,partner_key:int,business_unit_id:int,anon_cookie:chararray,machine_name:chararray,pagehost:chararray,filenameextension:chararray,referrerpath:chararray,referrerhost:chararray,referring_oid:chararray,referring_legacy_oid:chararray,ctype:chararray,cval:chararray,link_tag:chararray,link_type:chararray,sticky_tag:chararray,page_url:chararray,search_category:chararray,partner_subject:chararray,referring_partner_name:chararray,robot_pattern:chararray,browser:chararray,browser_major_version:chararray,browser_minor_version:chararray,os:chararray,os_family:chararray,ttag:chararray,dest_oid:chararray,global_id:chararray,hostname:chararray,path:chararray,filename:chararray,extension:chararray,query:chararray,user_agent:chararray,xrq:chararray,xref:chararray,page_guid:chararray,test_name:chararray,test_group:chararray,test_version:chararray,page_version:chararray,o_sticky_tag:chararray,new_referring_oid:chararray,day:chararray,network_ip:int,site_id:int,search_phrase:chararray,search_attributes:chararray,web_search_phrase:chararray,ip_address:chararray,is_pattern_match_robot:int,protocol:chararray,skc_title:chararray,skc_url:chararray,has_site_search_phrase:int,has_site_search_attribs:int,has_web_search_phrase:int,title_id:chararray,url_id:chararray,network_rev:int);

referrer_group_map = LOAD '/user/sgehlot/oozie/db_data/referrer_group_map'
as
(referrer_id:int, has_web_search_phrase:int, hostname:chararray,
referral_type_id:int,
referral_type_name:chararray,
referrer_group_id:int,referrer_group_name:chararray,referrer_group_cat_id:int,referrer_group_cat:chararray);

filter_pe = FILTER page_events BY is_iab_robot == 0 AND
is_pattern_match_robot == 0 AND day == '2011-05-10';

select_pe_col = FOREACH filter_pe GENERATE day, is_cookied_user,
anon_cookie, reg_cookie, referrer_id, has_web_search_phrase,
business_unit_id;

select_ref_col = FOREACH referrer_group_map GENERATE referrer_id,
has_web_search_phrase, referral_type_id;

jn = JOIN select_ref_col BY (referrer_id, has_web_search_phrase),
select_pe_col BY (referrer_id, has_web_search_phrase);
logic = FOREACH jn GENERATE
select_pe_col::day,
select_ref_col::referral_type_id,
 select_pe_col::business_unit_id,
((select_pe_col::is_cookied_user == 1) ? select_pe_col::anon_cookie : null)
as c_users,
 ((select_pe_col::is_cookied_user == 0) ? select_pe_col::anon_cookie : null)
as nc_users,
((select_pe_col::reg_cookie == '-1') ? null : select_pe_col::reg_cookie) as
registered_users;
 group_it = GROUP logic BY (select_pe_col::day,
   select_ref_col::referral_type_id,
   select_pe_col::business_unit_id);

agg_results = FOREACH group_it{
 dst_c_users = DISTINCT logic.c_users;
dst_nc_users = DISTINCT logic.nc_users;
 dst_registered_users = DISTINCT logic.registered_users;
      GENERATE
       group.select_pe_col::day,
        group.select_ref_col::referral_type_id,
group.select_pe_col::business_unit_id,
 COUNT(dst_c_users) as c_users,
COUNT(dst_nc_users) as nc_users,
COUNT(dst_registered_users) as registered_users;
 };

STORE agg_results INTO '/user/sgehlot/pt_users_referral_type_bu_day' USING
PigStorage('\t');
But I am keep on failing because of Java out of memory error. Data set in
"page_events" is huge but in "referrer_group_map" is relatively very less.

Here is error message:

Error: java.lang.OutOfMemoryError: GC overhead limit exceeded
at java.util.ArrayList.(ArrayList.java:112)
 at java.util.ArrayList.(ArrayList.java:119)
at org.apache.pig.data.DefaultTuple.(DefaultTuple.java:59)
 at org.apache.pig.data.BinSedesTuple.(BinSedesTuple.java:73)
at
org.apache.pig.data.BinSedesTupleFactory.newTuple(BinSedesTupleFactory.java:33)
 at
org.apache.pig.data.InternalCachedBag$CachedBagIterator.hasNext(InternalCachedBag.java:236)
at
org.apache.pig.builtin.Distinct.getDistinctFromNestedBags(Distinct.java:136)
 at org.apache.pig.builtin.Distinct.access$200(Distinct.java:38)
at org.apache.pig.builtin.Distinct$Intermediate.exec(Distinct.java:101)
 at org.apache.pig.builtin.Distinct$Intermediate.exec(Distinct.java:94)
at
org.apache.pig.backend.hadoop.executionengine.physicalLayer.expressionOperators.POUserFunc.getNext(POUserFunc.java:216)
 at
org.apache.pig.backend.hadoop.executionengine.physicalLayer.expressionOperators.POUserFunc.getNext(POUserFunc.java:253)
at
org.apache.pig.backend.hadoop.executionengine.physicalLayer.PhysicalOperator.getNext(PhysicalOperator.java:334)
 at
org.apache.pig.backend.hadoop.executionengine.physicalLayer.relationalOperators.POForEach.processPlan(POForEach.java:332)
at
org.apache.pig.backend.hadoop.executionengine.physicalLayer.relationalOperators.POForEach.getNext(POForEach.java:284)
 at
org.apache.pig.backend.hadoop.executionengine.physicalLayer.PhysicalOperator.processInput(PhysicalOperator.java:290)
at
org.apache.pig.backend.hadoop.executionengine.physicalLayer.relationalOperators.POLocalRearrange.getNext(POLocalRearrange.java:256)
 at
org.apache.pig.backend.hadoop.executionengine.mapReduce

The stack trace shows that the OOM error is happening when the distinct is
being applied. It looks like in some record(s) of the relation group_it, one
more of the following bags is very large - logic.c_users,  logic.nc_users or
logic.registered_users;

Try setting the property pig.cachedbag.memusage to 0.1 or lower (
-Dpig.cachedbag.memusage=0.1 on java command line). It controls the memory
used by pig internal bags, including those used by distinct.

If that does not work, you can try computing count-distinct for each type of
user separately and then combining the result.
You might want to have a look at this way of optimizing count-distinct
queries where skew can be a problem -
https://issues.apache.org/jira/browse/PIG-1846

-thejas

On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]> wrote:

> Hi Guys,
>
> I am running following Pig script in Pig 0.8 version
>
> page_events = LOAD '/user/sgehlot/day=2011-05-10' as
> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
> page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_nu
> mber:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:in
> t,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbo
> und_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_
> id:int,ontology_node_id:int,content_namespace_id:int,product_namespace_id:int,
> transparent_edition_id:int,default_edition_id:int,event_seq_num:int,is_last_pa
> ge:int,is_new_user:int,page_duration:int,page_seq_num:int,session_id:long,time
> _since_sess_start:int,reg_cookie:chararray,urs_app_id:int,is_reg_user:int,edit
> ion_id:int,user_agent_id:int,page_type_key:int,referrer_id:int,channel_id:int,
> level2_id:int,level3_id:int,brand_id:int,content_key:int,product_key:int,editi
> on_key:int,partner_key:int,business_unit_id:int,anon_cookie:chararray,machine_
> name:chararray,pagehost:chararray,filenameextension:chararray,referrerpath:cha
> rarray,referrerhost:chararray,referring_oid:chararray,referring_legacy_oid:cha
> rarray,ctype:chararray,cval:chararray,link_tag:chararray,link_type:chararray,s
> ticky_tag:chararray,page_url:chararray,search_category:chararray,partner_subje
> ct:chararray,referring_partner_name:chararray,robot_pattern:chararray,browser:
> chararray,browser_major_version:chararray,browser_minor_version:chararray,os:c
> hararray,os_family:chararray,ttag:chararray,dest_oid:chararray,global_id:chara
> rray,hostname:chararray,path:chararray,filename:chararray,extension:chararray,
> query:chararray,user_agent:chararray,xrq:chararray,xref:chararray,page_guid:ch
> ararray,test_name:chararray,test_group:chararray,test_version:chararray,page_v
> ersion:chararray,o_sticky_tag:chararray,new_referring_oid:chararray,day:charar
> ray,network_ip:int,site_id:int,search_phrase:chararray,search_attributes:chara
> rray,web_search_phrase:chararray,ip_address:chararray,is_pattern_match_robot:i
> nt,protocol:chararray,skc_title:chararray,skc_url:chararray,has_site_search_ph
> rase:int,has_site_search_attribs:int,has_web_search_phrase:int,title_id:charar
> ray,url_id:chararray,network_rev:int);
>
> referrer_group_map = LOAD '/user/sgehlot/oozie/db_data/referrer_group_map'
> as
> (referrer_id:int, has_web_search_phrase:int, hostname:chararray,
> referral_type_id:int,
> referral_type_name:chararray,
> referrer_group_id:int,referrer_group_name:chararray,referrer_group_cat_id:int,
> referrer_group_cat:chararray);
>
> filter_pe = FILTER page_events BY is_iab_robot == 0 AND
> is_pattern_match_robot == 0 AND day == '2011-05-10';
>
> select_pe_col = FOREACH filter_pe GENERATE day, is_cookied_user,
> anon_cookie, reg_cookie, referrer_id, has_web_search_phrase,
> business_unit_id;
>
> select_ref_col = FOREACH referrer_group_map GENERATE referrer_id,
> has_web_search_phrase, referral_type_id;
>
> jn = JOIN select_ref_col BY (referrer_id, has_web_search_phrase),
org.apache.pig.data.BinSedesTupleFactory.newTuple(BinSedesTupleFactory.java:33>
)

Hey Thejas,

I tried setting up property pig.cachedbag.memusage to 0.1 and also tried
computing distinct count for each type separately but still I am getting
errors like

Error: java.lang.OutOfMemoryError: Java heap space
Caused by: java.lang.OutOfMemoryError: GC overhead limit exceeded
java.io.IOException: Spill failed

Is there is some other way to do distinct - count that you may suggest?

Thanks for your help.

Sonia

On Fri, May 13, 2011 at 4:46 PM, Thejas M Nair <[EMAIL PROTECTED]> wrote:

> The stack trace shows that the OOM error is happening when the distinct is
> being applied. It looks like in some record(s) of the relation group_it,
> one
> more of the following bags is very large - logic.c_users,  logic.nc_users
> or
> logic.registered_users;
>
> Try setting the property pig.cachedbag.memusage to 0.1 or lower (
> -Dpig.cachedbag.memusage=0.1 on java command line). It controls the memory
> used by pig internal bags, including those used by distinct.
>
> If that does not work, you can try computing count-distinct for each type
> of
> user separately and then combining the result.
>
>
> You might want to have a look at this way of optimizing count-distinct
> queries where skew can be a problem -
> https://issues.apache.org/jira/browse/PIG-1846
>
> -thejas
>
>
>
>
>
> On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]> wrote:
>
> > Hi Guys,
> >
> > I am running following Pig script in Pig 0.8 version
> >
> > page_events = LOAD '/user/sgehlot/day=2011-05-10' as
> >
> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
> >
> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
> >
> page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_nu
> >
> mber:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:in
> >
> t,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbo
> >
> und_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_
> >
> id:int,ontology_node_id:int,content_namespace_id:int,product_namespace_id:int,
> >
> transparent_edition_id:int,default_edition_id:int,event_seq_num:int,is_last_pa
> >
> ge:int,is_new_user:int,page_duration:int,page_seq_num:int,session_id:long,time
> >
> _since_sess_start:int,reg_cookie:chararray,urs_app_id:int,is_reg_user:int,edit
> >
> ion_id:int,user_agent_id:int,page_type_key:int,referrer_id:int,channel_id:int,
> >
> level2_id:int,level3_id:int,brand_id:int,content_key:int,product_key:int,editi
> >
> on_key:int,partner_key:int,business_unit_id:int,anon_cookie:chararray,machine_
> >
> name:chararray,pagehost:chararray,filenameextension:chararray,referrerpath:cha
> >
> rarray,referrerhost:chararray,referring_oid:chararray,referring_legacy_oid:cha
> >
> rarray,ctype:chararray,cval:chararray,link_tag:chararray,link_type:chararray,s
> >
> ticky_tag:chararray,page_url:chararray,search_category:chararray,partner_subje
> >
> ct:chararray,referring_partner_name:chararray,robot_pattern:chararray,browser:
> >
> chararray,browser_major_version:chararray,browser_minor_version:chararray,os:c
> >
> hararray,os_family:chararray,ttag:chararray,dest_oid:chararray,global_id:chara
> >
> rray,hostname:chararray,path:chararray,filename:chararray,extension:chararray,
> >
> query:chararray,user_agent:chararray,xrq:chararray,xref:chararray,page_guid:ch
> >
> ararray,test_name:chararray,test_group:chararray,test_version:chararray,page_v
> >
> ersion:chararray,o_sticky_tag:chararray,new_referring_oid:chararray,day:charar
> >
> ray,network_ip:int,site_id:int,search_phrase:chararray,search_attributes:chara
> >
> rray,web_search_phrase:chararray,ip_address:chararray,is_pattern_match_robot:i
> >
> nt,protocol:chararray,skc_title:chararray,skc_url:chararray,has_site_search_ph
> >
> rase:int,has_site_search_attribs:int,has_web_search_phrase:int,title_id:charar
> > ray,url_id:chararray,network_rev:int);
> >
> > referrer_group_map = LOAD
> '/user/sgehlot/oozie/db_data/referrer_group_map'

you can try using some divide and conquer, like this:

a = group data by (key, SUBSTRING(the_field_to_be_distinct, 0, 2));
b = foreach a { x = distinct a.he_field_to_be_distinct; generate
group.key as key, COUNT(x) as cnt; }
c = group b by key;
d = foreach c generate group as key, SUM(b.cnt) as cnt;

using longer substring if still running into OOM.

Regards,
Shawn

On Fri, May 20, 2011 at 1:11 PM, sonia gehlot <[EMAIL PROTECTED]> wrote:
> Hey Thejas,
>
> I tried setting up property pig.cachedbag.memusage to 0.1 and also tried
> computing distinct count for each type separately but still I am getting
> errors like
>
> Error: java.lang.OutOfMemoryError: Java heap space
> Caused by: java.lang.OutOfMemoryError: GC overhead limit exceeded
> java.io.IOException: Spill failed
>
> Is there is some other way to do distinct - count that you may suggest?
>
> Thanks for your help.
>
> Sonia
>
> On Fri, May 13, 2011 at 4:46 PM, Thejas M Nair <[EMAIL PROTECTED]> wrote:
>
>> The stack trace shows that the OOM error is happening when the distinct is
>> being applied. It looks like in some record(s) of the relation group_it,
>> one
>> more of the following bags is very large - logic.c_users,  logic.nc_users
>> or
>> logic.registered_users;
>>
>> Try setting the property pig.cachedbag.memusage to 0.1 or lower (
>> -Dpig.cachedbag.memusage=0.1 on java command line). It controls the memory
>> used by pig internal bags, including those used by distinct.
>>
>> If that does not work, you can try computing count-distinct for each type
>> of
>> user separately and then combining the result.
>>
>>
>> You might want to have a look at this way of optimizing count-distinct
>> queries where skew can be a problem -
>> https://issues.apache.org/jira/browse/PIG-1846
>>
>> -thejas
>>
>>
>>
>>
>>
>> On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]> wrote:
>>
>> > Hi Guys,
>> >
>> > I am running following Pig script in Pig 0.8 version
>> >
>> > page_events = LOAD '/user/sgehlot/day=2011-05-10' as
>> >
>> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
>> >
>> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
>> >
>> page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_nu
>> >
>> mber:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:in
>> >
>> t,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbo
>> >
>> und_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_
>> >
>> id:int,ontology_node_id:int,content_namespace_id:int,product_namespace_id:int,
>> >
>> transparent_edition_id:int,default_edition_id:int,event_seq_num:int,is_last_pa
>> >
>> ge:int,is_new_user:int,page_duration:int,page_seq_num:int,session_id:long,time
>> >
>> _since_sess_start:int,reg_cookie:chararray,urs_app_id:int,is_reg_user:int,edit
>> >
>> ion_id:int,user_agent_id:int,page_type_key:int,referrer_id:int,channel_id:int,
>> >
>> level2_id:int,level3_id:int,brand_id:int,content_key:int,product_key:int,editi
>> >
>> on_key:int,partner_key:int,business_unit_id:int,anon_cookie:chararray,machine_
>> >
>> name:chararray,pagehost:chararray,filenameextension:chararray,referrerpath:cha
>> >
>> rarray,referrerhost:chararray,referring_oid:chararray,referring_legacy_oid:cha
>> >
>> rarray,ctype:chararray,cval:chararray,link_tag:chararray,link_type:chararray,s
>> >
>> ticky_tag:chararray,page_url:chararray,search_category:chararray,partner_subje
>> >
>> ct:chararray,referring_partner_name:chararray,robot_pattern:chararray,browser:
>> >
>> chararray,browser_major_version:chararray,browser_minor_version:chararray,os:c
>> >
>> hararray,os_family:chararray,ttag:chararray,dest_oid:chararray,global_id:chara
>> >
>> rray,hostname:chararray,path:chararray,filename:chararray,extension:chararray,
>> >
>> query:chararray,user_agent:chararray,xrq:chararray,xref:chararray,page_guid:ch
>> >
>> ararray,test_name:chararray,test_group:chararray,test_version:chararray,page_v

Hey, I am sorry but I din't get how substring will help in this?

On Fri, May 20, 2011 at 1:08 PM, Xiaomeng Wan <[EMAIL PROTECTED]> wrote:

> you can try using some divide and conquer, like this:
>
> a = group data by (key, SUBSTRING(the_field_to_be_distinct, 0, 2));
> b = foreach a { x = distinct a.he_field_to_be_distinct; generate
> group.key as key, COUNT(x) as cnt; }
> c = group b by key;
> d = foreach c generate group as key, SUM(b.cnt) as cnt;
>
> using longer substring if still running into OOM.
>
> Regards,
> Shawn
>
> On Fri, May 20, 2011 at 1:11 PM, sonia gehlot <[EMAIL PROTECTED]>
> wrote:
> > Hey Thejas,
> >
> > I tried setting up property pig.cachedbag.memusage to 0.1 and also tried
> > computing distinct count for each type separately but still I am getting
> > errors like
> >
> > Error: java.lang.OutOfMemoryError: Java heap space
> > Caused by: java.lang.OutOfMemoryError: GC overhead limit exceeded
> > java.io.IOException: Spill failed
> >
> > Is there is some other way to do distinct - count that you may suggest?
> >
> > Thanks for your help.
> >
> > Sonia
> >
> > On Fri, May 13, 2011 at 4:46 PM, Thejas M Nair <[EMAIL PROTECTED]>
> wrote:
> >
> >> The stack trace shows that the OOM error is happening when the distinct
> is
> >> being applied. It looks like in some record(s) of the relation group_it,
> >> one
> >> more of the following bags is very large - logic.c_users,
>  logic.nc_users
> >> or
> >> logic.registered_users;
> >>
> >> Try setting the property pig.cachedbag.memusage to 0.1 or lower (
> >> -Dpig.cachedbag.memusage=0.1 on java command line). It controls the
> memory
> >> used by pig internal bags, including those used by distinct.
> >>
> >> If that does not work, you can try computing count-distinct for each
> type
> >> of
> >> user separately and then combining the result.
> >>
> >>
> >> You might want to have a look at this way of optimizing count-distinct
> >> queries where skew can be a problem -
> >> https://issues.apache.org/jira/browse/PIG-1846
> >>
> >> -thejas
> >>
> >>
> >>
> >>
> >>
> >> On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]> wrote:
> >>
> >> > Hi Guys,
> >> >
> >> > I am running following Pig script in Pig 0.8 version
> >> >
> >> > page_events = LOAD '/user/sgehlot/day=2011-05-10' as
> >> >
> >>
> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
> >> >
> >>
> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
> >> >
> >>
> page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_nu
> >> >
> >>
> mber:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:in
> >> >
> >>
> t,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbo
> >> >
> >>
> und_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_
> >> >
> >>
> id:int,ontology_node_id:int,content_namespace_id:int,product_namespace_id:int,
> >> >
> >>
> transparent_edition_id:int,default_edition_id:int,event_seq_num:int,is_last_pa
> >> >
> >>
> ge:int,is_new_user:int,page_duration:int,page_seq_num:int,session_id:long,time
> >> >
> >>
> _since_sess_start:int,reg_cookie:chararray,urs_app_id:int,is_reg_user:int,edit
> >> >
> >>
> ion_id:int,user_agent_id:int,page_type_key:int,referrer_id:int,channel_id:int,
> >> >
> >>
> level2_id:int,level3_id:int,brand_id:int,content_key:int,product_key:int,editi
> >> >
> >>
> on_key:int,partner_key:int,business_unit_id:int,anon_cookie:chararray,machine_
> >> >
> >>
> name:chararray,pagehost:chararray,filenameextension:chararray,referrerpath:cha
> >> >
> >>
> rarray,referrerhost:chararray,referring_oid:chararray,referring_legacy_oid:cha
> >> >
> >>
> rarray,ctype:chararray,cval:chararray,link_tag:chararray,link_type:chararray,s
> >> >
> >>
> ticky_tag:chararray,page_url:chararray,search_category:chararray,partner_subje
> >> >
> >>
> ct:chararray,referring_partner_name:chararray,robot_pattern:chararray,browser:
> >> >
> >>
> chararray,browser_major_version:chararray,browser_minor_version:chararray,os:c

It servers two purposes:
1. divide the group into smaller subgroups
2. make sure distinct in subgroup => distinct in group

Shawn

On Fri, May 20, 2011 at 2:20 PM, sonia gehlot <[EMAIL PROTECTED]> wrote:
> Hey, I am sorry but I din't get how substring will help in this?
>
> On Fri, May 20, 2011 at 1:08 PM, Xiaomeng Wan <[EMAIL PROTECTED]> wrote:
>
>> you can try using some divide and conquer, like this:
>>
>> a = group data by (key, SUBSTRING(the_field_to_be_distinct, 0, 2));
>> b = foreach a { x = distinct a.he_field_to_be_distinct; generate
>> group.key as key, COUNT(x) as cnt; }
>> c = group b by key;
>> d = foreach c generate group as key, SUM(b.cnt) as cnt;
>>
>> using longer substring if still running into OOM.
>>
>> Regards,
>> Shawn
>>
>> On Fri, May 20, 2011 at 1:11 PM, sonia gehlot <[EMAIL PROTECTED]>
>> wrote:
>> > Hey Thejas,
>> >
>> > I tried setting up property pig.cachedbag.memusage to 0.1 and also tried
>> > computing distinct count for each type separately but still I am getting
>> > errors like
>> >
>> > Error: java.lang.OutOfMemoryError: Java heap space
>> > Caused by: java.lang.OutOfMemoryError: GC overhead limit exceeded
>> > java.io.IOException: Spill failed
>> >
>> > Is there is some other way to do distinct - count that you may suggest?
>> >
>> > Thanks for your help.
>> >
>> > Sonia
>> >
>> > On Fri, May 13, 2011 at 4:46 PM, Thejas M Nair <[EMAIL PROTECTED]>
>> wrote:
>> >
>> >> The stack trace shows that the OOM error is happening when the distinct
>> is
>> >> being applied. It looks like in some record(s) of the relation group_it,
>> >> one
>> >> more of the following bags is very large - logic.c_users,
>>  logic.nc_users
>> >> or
>> >> logic.registered_users;
>> >>
>> >> Try setting the property pig.cachedbag.memusage to 0.1 or lower (
>> >> -Dpig.cachedbag.memusage=0.1 on java command line). It controls the
>> memory
>> >> used by pig internal bags, including those used by distinct.
>> >>
>> >> If that does not work, you can try computing count-distinct for each
>> type
>> >> of
>> >> user separately and then combining the result.
>> >>
>> >>
>> >> You might want to have a look at this way of optimizing count-distinct
>> >> queries where skew can be a problem -
>> >> https://issues.apache.org/jira/browse/PIG-1846
>> >>
>> >> -thejas
>> >>
>> >>
>> >>
>> >>
>> >>
>> >> On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]> wrote:
>> >>
>> >> > Hi Guys,
>> >> >
>> >> > I am running following Pig script in Pig 0.8 version
>> >> >
>> >> > page_events = LOAD '/user/sgehlot/day=2011-05-10' as
>> >> >
>> >>
>> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
>> >> >
>> >>
>> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
>> >> >
>> >>
>> page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_nu
>> >> >
>> >>
>> mber:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:in
>> >> >
>> >>
>> t,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbo
>> >> >
>> >>
>> und_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_
>> >> >
>> >>
>> id:int,ontology_node_id:int,content_namespace_id:int,product_namespace_id:int,
>> >> >
>> >>
>> transparent_edition_id:int,default_edition_id:int,event_seq_num:int,is_last_pa
>> >> >
>> >>
>> ge:int,is_new_user:int,page_duration:int,page_seq_num:int,session_id:long,time
>> >> >
>> >>
>> _since_sess_start:int,reg_cookie:chararray,urs_app_id:int,is_reg_user:int,edit
>> >> >
>> >>
>> ion_id:int,user_agent_id:int,page_type_key:int,referrer_id:int,channel_id:int,
>> >> >
>> >>
>> level2_id:int,level3_id:int,brand_id:int,content_key:int,product_key:int,editi
>> >> >
>> >>
>> on_key:int,partner_key:int,business_unit_id:int,anon_cookie:chararray,machine_
>> >> >
>> >>
>> name:chararray,pagehost:chararray,filenameextension:chararray,referrerpath:cha
>> >> >
>> >>
>> rarray,referrerhost:chararray,referring_oid:chararray,referring_legacy_oid:cha

Hi Shawn,

I tried using SUBSTRING in my script with different combinations but still
getting OOM errors.

is there is any other alternative to use distinct - count against very large
set of data.

Thanks,
Sonia

On Fri, May 20, 2011 at 1:54 PM, Xiaomeng Wan <[EMAIL PROTECTED]> wrote:

> It servers two purposes:
> 1. divide the group into smaller subgroups
> 2. make sure distinct in subgroup => distinct in group
>
> Shawn
>
> On Fri, May 20, 2011 at 2:20 PM, sonia gehlot <[EMAIL PROTECTED]>
> wrote:
> > Hey, I am sorry but I din't get how substring will help in this?
> >
> > On Fri, May 20, 2011 at 1:08 PM, Xiaomeng Wan <[EMAIL PROTECTED]>
> wrote:
> >
> >> you can try using some divide and conquer, like this:
> >>
> >> a = group data by (key, SUBSTRING(the_field_to_be_distinct, 0, 2));
> >> b = foreach a { x = distinct a.he_field_to_be_distinct; generate
> >> group.key as key, COUNT(x) as cnt; }
> >> c = group b by key;
> >> d = foreach c generate group as key, SUM(b.cnt) as cnt;
> >>
> >> using longer substring if still running into OOM.
> >>
> >> Regards,
> >> Shawn
> >>
> >> On Fri, May 20, 2011 at 1:11 PM, sonia gehlot <[EMAIL PROTECTED]>
> >> wrote:
> >> > Hey Thejas,
> >> >
> >> > I tried setting up property pig.cachedbag.memusage to 0.1 and also
> tried
> >> > computing distinct count for each type separately but still I am
> getting
> >> > errors like
> >> >
> >> > Error: java.lang.OutOfMemoryError: Java heap space
> >> > Caused by: java.lang.OutOfMemoryError: GC overhead limit exceeded
> >> > java.io.IOException: Spill failed
> >> >
> >> > Is there is some other way to do distinct - count that you may
> suggest?
> >> >
> >> > Thanks for your help.
> >> >
> >> > Sonia
> >> >
> >> > On Fri, May 13, 2011 at 4:46 PM, Thejas M Nair <[EMAIL PROTECTED]>
> >> wrote:
> >> >
> >> >> The stack trace shows that the OOM error is happening when the
> distinct
> >> is
> >> >> being applied. It looks like in some record(s) of the relation
> group_it,
> >> >> one
> >> >> more of the following bags is very large - logic.c_users,
> >>  logic.nc_users
> >> >> or
> >> >> logic.registered_users;
> >> >>
> >> >> Try setting the property pig.cachedbag.memusage to 0.1 or lower (
> >> >> -Dpig.cachedbag.memusage=0.1 on java command line). It controls the
> >> memory
> >> >> used by pig internal bags, including those used by distinct.
> >> >>
> >> >> If that does not work, you can try computing count-distinct for each
> >> type
> >> >> of
> >> >> user separately and then combining the result.
> >> >>
> >> >>
> >> >> You might want to have a look at this way of optimizing
> count-distinct
> >> >> queries where skew can be a problem -
> >> >> https://issues.apache.org/jira/browse/PIG-1846
> >> >>
> >> >> -thejas
> >> >>
> >> >>
> >> >>
> >> >>
> >> >>
> >> >> On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]> wrote:
> >> >>
> >> >> > Hi Guys,
> >> >> >
> >> >> > I am running following Pig script in Pig 0.8 version
> >> >> >
> >> >> > page_events = LOAD '/user/sgehlot/day=2011-05-10' as
> >> >> >
> >> >>
> >>
> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
> >> >> >
> >> >>
> >>
> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
> >> >> >
> >> >>
> >>
> page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_nu
> >> >> >
> >> >>
> >>
> mber:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:in
> >> >> >
> >> >>
> >>
> t,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbo
> >> >> >
> >> >>
> >>
> und_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_
> >> >> >
> >> >>
> >>
> id:int,ontology_node_id:int,content_namespace_id:int,product_namespace_id:int,
> >> >> >
> >> >>
> >>
> transparent_edition_id:int,default_edition_id:int,event_seq_num:int,is_last_pa
> >> >> >
> >> >>
> >>
> ge:int,is_new_user:int,page_duration:int,page_seq_num:int,session_id:long,time
> >> >> >
> >> >>

you can group by your key + the thing you want distinct counts of, and
generate counts of those.

On Mon, May 23, 2011 at 2:17 PM, sonia gehlot <[EMAIL PROTECTED]> wrote:
> Hi Shawn,
>
> I tried using SUBSTRING in my script with different combinations but still
> getting OOM errors.
>
> is there is any other alternative to use distinct - count against very large
> set of data.
>
> Thanks,
> Sonia
>
> On Fri, May 20, 2011 at 1:54 PM, Xiaomeng Wan <[EMAIL PROTECTED]> wrote:
>
>> It servers two purposes:
>> 1. divide the group into smaller subgroups
>> 2. make sure distinct in subgroup => distinct in group
>>
>> Shawn
>>
>> On Fri, May 20, 2011 at 2:20 PM, sonia gehlot <[EMAIL PROTECTED]>
>> wrote:
>> > Hey, I am sorry but I din't get how substring will help in this?
>> >
>> > On Fri, May 20, 2011 at 1:08 PM, Xiaomeng Wan <[EMAIL PROTECTED]>
>> wrote:
>> >
>> >> you can try using some divide and conquer, like this:
>> >>
>> >> a = group data by (key, SUBSTRING(the_field_to_be_distinct, 0, 2));
>> >> b = foreach a { x = distinct a.he_field_to_be_distinct; generate
>> >> group.key as key, COUNT(x) as cnt; }
>> >> c = group b by key;
>> >> d = foreach c generate group as key, SUM(b.cnt) as cnt;
>> >>
>> >> using longer substring if still running into OOM.
>> >>
>> >> Regards,
>> >> Shawn
>> >>
>> >> On Fri, May 20, 2011 at 1:11 PM, sonia gehlot <[EMAIL PROTECTED]>
>> >> wrote:
>> >> > Hey Thejas,
>> >> >
>> >> > I tried setting up property pig.cachedbag.memusage to 0.1 and also
>> tried
>> >> > computing distinct count for each type separately but still I am
>> getting
>> >> > errors like
>> >> >
>> >> > Error: java.lang.OutOfMemoryError: Java heap space
>> >> > Caused by: java.lang.OutOfMemoryError: GC overhead limit exceeded
>> >> > java.io.IOException: Spill failed
>> >> >
>> >> > Is there is some other way to do distinct - count that you may
>> suggest?
>> >> >
>> >> > Thanks for your help.
>> >> >
>> >> > Sonia
>> >> >
>> >> > On Fri, May 13, 2011 at 4:46 PM, Thejas M Nair <[EMAIL PROTECTED]>
>> >> wrote:
>> >> >
>> >> >> The stack trace shows that the OOM error is happening when the
>> distinct
>> >> is
>> >> >> being applied. It looks like in some record(s) of the relation
>> group_it,
>> >> >> one
>> >> >> more of the following bags is very large - logic.c_users,
>> >>  logic.nc_users
>> >> >> or
>> >> >> logic.registered_users;
>> >> >>
>> >> >> Try setting the property pig.cachedbag.memusage to 0.1 or lower (
>> >> >> -Dpig.cachedbag.memusage=0.1 on java command line). It controls the
>> >> memory
>> >> >> used by pig internal bags, including those used by distinct.
>> >> >>
>> >> >> If that does not work, you can try computing count-distinct for each
>> >> type
>> >> >> of
>> >> >> user separately and then combining the result.
>> >> >>
>> >> >>
>> >> >> You might want to have a look at this way of optimizing
>> count-distinct
>> >> >> queries where skew can be a problem -
>> >> >> https://issues.apache.org/jira/browse/PIG-1846
>> >> >>
>> >> >> -thejas
>> >> >>
>> >> >>
>> >> >>
>> >> >>
>> >> >>
>> >> >> On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]> wrote:
>> >> >>
>> >> >> > Hi Guys,
>> >> >> >
>> >> >> > I am running following Pig script in Pig 0.8 version
>> >> >> >
>> >> >> > page_events = LOAD '/user/sgehlot/day=2011-05-10' as
>> >> >> >
>> >> >>
>> >>
>> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
>> >> >> >
>> >> >>
>> >>
>> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,
>> >> >> >
>> >> >>
>> >>
>> page_type_id:int,content_id:int,product_id:int,referrer_edition_id:int,page_nu
>> >> >> >
>> >> >>
>> >>
>> mber:int,is_iab_robot:int,browser_id:int,os_id:int,dw_pubsys_id:int,refresh:in
>> >> >> >
>> >> >>
>> >>
>> t,asset_id:int,asset_type_id:int,content_type_id:int,product_type_id:int,outbo
>> >> >> >
>> >> >>
>> >>
>> und_email_id:long,gbal_clc:int,mtype:int,user_action_id:int,referring_partner_
>> >> >> >
>> >> >>

Hi All,

This script worked for me by setting following condition

*set pig.exec.nocombiner true;*

Thanks for your help.

Sonia

On Mon, May 23, 2011 at 3:03 PM, Dmitriy Ryaboy <[EMAIL PROTECTED]> wrote:

> you can group by your key + the thing you want distinct counts of, and
> generate counts of those.
>
> On Mon, May 23, 2011 at 2:17 PM, sonia gehlot <[EMAIL PROTECTED]>
> wrote:
> > Hi Shawn,
> >
> > I tried using SUBSTRING in my script with different combinations but
> still
> > getting OOM errors.
> >
> > is there is any other alternative to use distinct - count against very
> large
> > set of data.
> >
> > Thanks,
> > Sonia
> >
> > On Fri, May 20, 2011 at 1:54 PM, Xiaomeng Wan <[EMAIL PROTECTED]>
> wrote:
> >
> >> It servers two purposes:
> >> 1. divide the group into smaller subgroups
> >> 2. make sure distinct in subgroup => distinct in group
> >>
> >> Shawn
> >>
> >> On Fri, May 20, 2011 at 2:20 PM, sonia gehlot <[EMAIL PROTECTED]>
> >> wrote:
> >> > Hey, I am sorry but I din't get how substring will help in this?
> >> >
> >> > On Fri, May 20, 2011 at 1:08 PM, Xiaomeng Wan <[EMAIL PROTECTED]>
> >> wrote:
> >> >
> >> >> you can try using some divide and conquer, like this:
> >> >>
> >> >> a = group data by (key, SUBSTRING(the_field_to_be_distinct, 0, 2));
> >> >> b = foreach a { x = distinct a.he_field_to_be_distinct; generate
> >> >> group.key as key, COUNT(x) as cnt; }
> >> >> c = group b by key;
> >> >> d = foreach c generate group as key, SUM(b.cnt) as cnt;
> >> >>
> >> >> using longer substring if still running into OOM.
> >> >>
> >> >> Regards,
> >> >> Shawn
> >> >>
> >> >> On Fri, May 20, 2011 at 1:11 PM, sonia gehlot <
> [EMAIL PROTECTED]>
> >> >> wrote:
> >> >> > Hey Thejas,
> >> >> >
> >> >> > I tried setting up property pig.cachedbag.memusage to 0.1 and also
> >> tried
> >> >> > computing distinct count for each type separately but still I am
> >> getting
> >> >> > errors like
> >> >> >
> >> >> > Error: java.lang.OutOfMemoryError: Java heap space
> >> >> > Caused by: java.lang.OutOfMemoryError: GC overhead limit exceeded
> >> >> > java.io.IOException: Spill failed
> >> >> >
> >> >> > Is there is some other way to do distinct - count that you may
> >> suggest?
> >> >> >
> >> >> > Thanks for your help.
> >> >> >
> >> >> > Sonia
> >> >> >
> >> >> > On Fri, May 13, 2011 at 4:46 PM, Thejas M Nair <
> [EMAIL PROTECTED]>
> >> >> wrote:
> >> >> >
> >> >> >> The stack trace shows that the OOM error is happening when the
> >> distinct
> >> >> is
> >> >> >> being applied. It looks like in some record(s) of the relation
> >> group_it,
> >> >> >> one
> >> >> >> more of the following bags is very large - logic.c_users,
> >> >>  logic.nc_users
> >> >> >> or
> >> >> >> logic.registered_users;
> >> >> >>
> >> >> >> Try setting the property pig.cachedbag.memusage to 0.1 or lower (
> >> >> >> -Dpig.cachedbag.memusage=0.1 on java command line). It controls
> the
> >> >> memory
> >> >> >> used by pig internal bags, including those used by distinct.
> >> >> >>
> >> >> >> If that does not work, you can try computing count-distinct for
> each
> >> >> type
> >> >> >> of
> >> >> >> user separately and then combining the result.
> >> >> >>
> >> >> >>
> >> >> >> You might want to have a look at this way of optimizing
> >> count-distinct
> >> >> >> queries where skew can be a problem -
> >> >> >> https://issues.apache.org/jira/browse/PIG-1846
> >> >> >>
> >> >> >> -thejas
> >> >> >>
> >> >> >>
> >> >> >>
> >> >> >>
> >> >> >>
> >> >> >> On 5/12/11 10:43 AM, "sonia gehlot" <[EMAIL PROTECTED]>
> wrote:
> >> >> >>
> >> >> >> > Hi Guys,
> >> >> >> >
> >> >> >> > I am running following Pig script in Pig 0.8 version
> >> >> >> >
> >> >> >> > page_events = LOAD '/user/sgehlot/day=2011-05-10' as
> >> >> >> >
> >> >> >>
> >> >>
> >>
> (event_dt_ht:chararray,event_dt_ut:chararray,event_rec_num:int,event_type:int,
> >> >> >> >
> >> >> >>
> >> >>
> >>
> client_ip_addr:long,hub_id:int,is_cookied_user:int,local_ontology_node_id:int,