Re: [Dhis2-devs] Import of Events is extremely slow - any plans to improve it?
Hi Calle Depending on the configuration of the events (how many org units, data elements etc), yes, event import can still feel a bit slow. We did some updates in 225/226, but as you are already using 226 this didn't help your case. Yes. when we did the metadata importer we managed to speedup it a lot, but this required a complete rewrite, and to have the same gains in event import we might have to do the same (which means it will have to wait a while). Feel free to write a issue for it. -- Morten Olav Hansen Senior Engineer, DHIS 2 University of Oslo http://www.dhis2.org On Sat, Jul 1, 2017 at 2:20 PM, Calle Hedbergwrote: > David, > > I'm importing into 2.26, postgresql is tuned as far as possible (the test > was on my laptop), and 12GB RAM. I was using CSV, put there should not be > any significant difference between JSON and CSV. > > I will import the 2.5 mill events directly and see what time that takes, > comparatively. > > Regards > Calle > > On 30 June 2017 at 18:05, David Huser wrote: > >> Hi Calle >> >> When some conditions are met, the event import can be really speedy >> (imports of ~100 000 in <30 minutes). >> >> - Recent version (2.25+) >> - PostgreSQL properly tuned (this is important) >> - Enough RAM (8 GB+) >> >> I usually use JSON files and post them against /api/events. >> >> David >> >> On Fri, Jun 30, 2017 at 12:58 PM, Calle Hedberg >> wrote: >> >>> Ime, >>> >>> I was using the UI - have not looked at the API endpoint. >>> >>> For now I will use sql scripts as I've done before - I'm just trying to >>> find out if anybody is working on improving what is (now) an import >>> function that in reality only caters for small data sets. There is no >>> question about the core developers ability to speed this up - I saw that >>> with the import of OU meta-data - but not sure if anybody has done anything. >>> >>> If no answers from core developers, the I will write a JIRA issue for it. >>> >>> Regards >>> Calle >>> >>> On 29 June 2017 at 02:25, Ime Asangansi wrote: >>> Hi Calle, Thanks for your question. Just curious (I don't have answer). As we are planning to do this soon. Are you doing this through UI or API? If API, which endpoint enables this? Thanks Ime On Jun 28, 2017 19:15, "Calle Hedberg" wrote: > Hi > > I started importing around 700,000 events 2 days ago (about 5 mill > individual values) - and the import is still running, 48 hours later. > Import is slowing - it seemed to be importing around 7-8 values per > second, > now down to 2-3 per second. It looks like that import might take around > 200 > hours - which makes no sense from a usability point of view. Importing > OrgUnits used to be equally slow, but a revamp recently change a 10,000 > OrgUnit import from hours to something like 5 seconds. > > It seems obvious that event imports have been designed to import > values one by one, which is equivalent to asking people to fetch water > using a leaking tea-spoon. > > I've searched JIRA but cannot find any issue related to this, thus > this post: Are there any plans underway to make import of Event data > usable > for larger record sets? > > Regards > Calle > > *** > > Calle Hedberg > > 46D Alma Road, 7700 Rosebank, SOUTH AFRICA > > Tel/fax (home): +27-21-685-6472 <+27%2021%20685%206472> > > Cell: +27-82-853-5352 <+27%2082%20853%205352> > > Iridium SatPhone: +8816-315-19119 <+881%206%20315%2019119> > > Email: calle.hedb...@gmail.com > > Skype: calle_hedberg > > *** > > > ___ > Mailing list: https://launchpad.net/~dhis2-devs > Post to : dhis2-devs@lists.launchpad.net > Unsubscribe : https://launchpad.net/~dhis2-devs > More help : https://help.launchpad.net/ListHelp > > >>> >>> >>> -- >>> >>> *** >>> >>> Calle Hedberg >>> >>> 46D Alma Road, 7700 Rosebank, SOUTH AFRICA >>> >>> Tel/fax (home): +27-21-685-6472 <+27%2021%20685%206472> >>> >>> Cell: +27-82-853-5352 <+27%2082%20853%205352> >>> >>> Iridium SatPhone: +8816-315-19119 <+881%206%20315%2019119> >>> >>> Email: calle.hedb...@gmail.com >>> >>> Skype: calle_hedberg >>> >>> *** >>> >>> >>> ___ >>> Mailing list: https://launchpad.net/~dhis2-devs >>> Post to : dhis2-devs@lists.launchpad.net >>> Unsubscribe : https://launchpad.net/~dhis2-devs >>> More help : https://help.launchpad.net/ListHelp >>> >>> >> >> >> -- >> >> >> *David Huser* >> DHIS2 Support Specialist >>
Re: [Dhis2-devs] Import of Events is extremely slow - any plans to improve it?
David, I'm importing into 2.26, postgresql is tuned as far as possible (the test was on my laptop), and 12GB RAM. I was using CSV, put there should not be any significant difference between JSON and CSV. I will import the 2.5 mill events directly and see what time that takes, comparatively. Regards Calle On 30 June 2017 at 18:05, David Huserwrote: > Hi Calle > > When some conditions are met, the event import can be really speedy > (imports of ~100 000 in <30 minutes). > > - Recent version (2.25+) > - PostgreSQL properly tuned (this is important) > - Enough RAM (8 GB+) > > I usually use JSON files and post them against /api/events. > > David > > On Fri, Jun 30, 2017 at 12:58 PM, Calle Hedberg > wrote: > >> Ime, >> >> I was using the UI - have not looked at the API endpoint. >> >> For now I will use sql scripts as I've done before - I'm just trying to >> find out if anybody is working on improving what is (now) an import >> function that in reality only caters for small data sets. There is no >> question about the core developers ability to speed this up - I saw that >> with the import of OU meta-data - but not sure if anybody has done anything. >> >> If no answers from core developers, the I will write a JIRA issue for it. >> >> Regards >> Calle >> >> On 29 June 2017 at 02:25, Ime Asangansi wrote: >> >>> Hi Calle, >>> >>> Thanks for your question. >>> Just curious (I don't have answer). As we are planning to do this soon. >>> >>> Are you doing this through UI or API? >>> If API, which endpoint enables this? >>> >>> Thanks >>> >>> Ime >>> >>> >>> >>> On Jun 28, 2017 19:15, "Calle Hedberg" wrote: >>> Hi I started importing around 700,000 events 2 days ago (about 5 mill individual values) - and the import is still running, 48 hours later. Import is slowing - it seemed to be importing around 7-8 values per second, now down to 2-3 per second. It looks like that import might take around 200 hours - which makes no sense from a usability point of view. Importing OrgUnits used to be equally slow, but a revamp recently change a 10,000 OrgUnit import from hours to something like 5 seconds. It seems obvious that event imports have been designed to import values one by one, which is equivalent to asking people to fetch water using a leaking tea-spoon. I've searched JIRA but cannot find any issue related to this, thus this post: Are there any plans underway to make import of Event data usable for larger record sets? Regards Calle *** Calle Hedberg 46D Alma Road, 7700 Rosebank, SOUTH AFRICA Tel/fax (home): +27-21-685-6472 <+27%2021%20685%206472> Cell: +27-82-853-5352 <+27%2082%20853%205352> Iridium SatPhone: +8816-315-19119 <+881%206%20315%2019119> Email: calle.hedb...@gmail.com Skype: calle_hedberg *** ___ Mailing list: https://launchpad.net/~dhis2-devs Post to : dhis2-devs@lists.launchpad.net Unsubscribe : https://launchpad.net/~dhis2-devs More help : https://help.launchpad.net/ListHelp >> >> >> -- >> >> *** >> >> Calle Hedberg >> >> 46D Alma Road, 7700 Rosebank, SOUTH AFRICA >> >> Tel/fax (home): +27-21-685-6472 <+27%2021%20685%206472> >> >> Cell: +27-82-853-5352 <+27%2082%20853%205352> >> >> Iridium SatPhone: +8816-315-19119 <+881%206%20315%2019119> >> >> Email: calle.hedb...@gmail.com >> >> Skype: calle_hedberg >> >> *** >> >> >> ___ >> Mailing list: https://launchpad.net/~dhis2-devs >> Post to : dhis2-devs@lists.launchpad.net >> Unsubscribe : https://launchpad.net/~dhis2-devs >> More help : https://help.launchpad.net/ListHelp >> >> > > > -- > > > *David Huser* > DHIS2 Support Specialist > dhu...@baosystems.com | https://baosystems.com | Skype: dafhus | 2900 K > Street, Suite 406, Washington D.C. 20007 > -- *** Calle Hedberg 46D Alma Road, 7700 Rosebank, SOUTH AFRICA Tel/fax (home): +27-21-685-6472 Cell: +27-82-853-5352 Iridium SatPhone: +8816-315-19119 Email: calle.hedb...@gmail.com Skype: calle_hedberg *** ___ Mailing list: https://launchpad.net/~dhis2-devs Post to : dhis2-devs@lists.launchpad.net Unsubscribe : https://launchpad.net/~dhis2-devs More help : https://help.launchpad.net/ListHelp
Re: [Dhis2-devs] Import of Events is extremely slow - any plans to improve it?
Hi Calle When some conditions are met, the event import can be really speedy (imports of ~100 000 in <30 minutes). - Recent version (2.25+) - PostgreSQL properly tuned (this is important) - Enough RAM (8 GB+) I usually use JSON files and post them against /api/events. David On Fri, Jun 30, 2017 at 12:58 PM, Calle Hedbergwrote: > Ime, > > I was using the UI - have not looked at the API endpoint. > > For now I will use sql scripts as I've done before - I'm just trying to > find out if anybody is working on improving what is (now) an import > function that in reality only caters for small data sets. There is no > question about the core developers ability to speed this up - I saw that > with the import of OU meta-data - but not sure if anybody has done anything. > > If no answers from core developers, the I will write a JIRA issue for it. > > Regards > Calle > > On 29 June 2017 at 02:25, Ime Asangansi wrote: > >> Hi Calle, >> >> Thanks for your question. >> Just curious (I don't have answer). As we are planning to do this soon. >> >> Are you doing this through UI or API? >> If API, which endpoint enables this? >> >> Thanks >> >> Ime >> >> >> >> On Jun 28, 2017 19:15, "Calle Hedberg" wrote: >> >>> Hi >>> >>> I started importing around 700,000 events 2 days ago (about 5 mill >>> individual values) - and the import is still running, 48 hours later. >>> Import is slowing - it seemed to be importing around 7-8 values per second, >>> now down to 2-3 per second. It looks like that import might take around 200 >>> hours - which makes no sense from a usability point of view. Importing >>> OrgUnits used to be equally slow, but a revamp recently change a 10,000 >>> OrgUnit import from hours to something like 5 seconds. >>> >>> It seems obvious that event imports have been designed to import values >>> one by one, which is equivalent to asking people to fetch water using a >>> leaking tea-spoon. >>> >>> I've searched JIRA but cannot find any issue related to this, thus this >>> post: Are there any plans underway to make import of Event data usable for >>> larger record sets? >>> >>> Regards >>> Calle >>> >>> *** >>> >>> Calle Hedberg >>> >>> 46D Alma Road, 7700 Rosebank, SOUTH AFRICA >>> >>> Tel/fax (home): +27-21-685-6472 <+27%2021%20685%206472> >>> >>> Cell: +27-82-853-5352 <+27%2082%20853%205352> >>> >>> Iridium SatPhone: +8816-315-19119 <+881%206%20315%2019119> >>> >>> Email: calle.hedb...@gmail.com >>> >>> Skype: calle_hedberg >>> >>> *** >>> >>> >>> ___ >>> Mailing list: https://launchpad.net/~dhis2-devs >>> Post to : dhis2-devs@lists.launchpad.net >>> Unsubscribe : https://launchpad.net/~dhis2-devs >>> More help : https://help.launchpad.net/ListHelp >>> >>> > > > -- > > *** > > Calle Hedberg > > 46D Alma Road, 7700 Rosebank, SOUTH AFRICA > > Tel/fax (home): +27-21-685-6472 <+27%2021%20685%206472> > > Cell: +27-82-853-5352 <+27%2082%20853%205352> > > Iridium SatPhone: +8816-315-19119 <+881%206%20315%2019119> > > Email: calle.hedb...@gmail.com > > Skype: calle_hedberg > > *** > > > ___ > Mailing list: https://launchpad.net/~dhis2-devs > Post to : dhis2-devs@lists.launchpad.net > Unsubscribe : https://launchpad.net/~dhis2-devs > More help : https://help.launchpad.net/ListHelp > > -- *David Huser* DHIS2 Support Specialist dhu...@baosystems.com | https://baosystems.com | Skype: dafhus | 2900 K Street, Suite 406, Washington D.C. 20007 ___ Mailing list: https://launchpad.net/~dhis2-devs Post to : dhis2-devs@lists.launchpad.net Unsubscribe : https://launchpad.net/~dhis2-devs More help : https://help.launchpad.net/ListHelp
Re: [Dhis2-devs] Import of Events is extremely slow - any plans to improve it?
Ime, I was using the UI - have not looked at the API endpoint. For now I will use sql scripts as I've done before - I'm just trying to find out if anybody is working on improving what is (now) an import function that in reality only caters for small data sets. There is no question about the core developers ability to speed this up - I saw that with the import of OU meta-data - but not sure if anybody has done anything. If no answers from core developers, the I will write a JIRA issue for it. Regards Calle On 29 June 2017 at 02:25, Ime Asangansiwrote: > Hi Calle, > > Thanks for your question. > Just curious (I don't have answer). As we are planning to do this soon. > > Are you doing this through UI or API? > If API, which endpoint enables this? > > Thanks > > Ime > > > > On Jun 28, 2017 19:15, "Calle Hedberg" wrote: > >> Hi >> >> I started importing around 700,000 events 2 days ago (about 5 mill >> individual values) - and the import is still running, 48 hours later. >> Import is slowing - it seemed to be importing around 7-8 values per second, >> now down to 2-3 per second. It looks like that import might take around 200 >> hours - which makes no sense from a usability point of view. Importing >> OrgUnits used to be equally slow, but a revamp recently change a 10,000 >> OrgUnit import from hours to something like 5 seconds. >> >> It seems obvious that event imports have been designed to import values >> one by one, which is equivalent to asking people to fetch water using a >> leaking tea-spoon. >> >> I've searched JIRA but cannot find any issue related to this, thus this >> post: Are there any plans underway to make import of Event data usable for >> larger record sets? >> >> Regards >> Calle >> >> *** >> >> Calle Hedberg >> >> 46D Alma Road, 7700 Rosebank, SOUTH AFRICA >> >> Tel/fax (home): +27-21-685-6472 <+27%2021%20685%206472> >> >> Cell: +27-82-853-5352 <+27%2082%20853%205352> >> >> Iridium SatPhone: +8816-315-19119 <+881%206%20315%2019119> >> >> Email: calle.hedb...@gmail.com >> >> Skype: calle_hedberg >> >> *** >> >> >> ___ >> Mailing list: https://launchpad.net/~dhis2-devs >> Post to : dhis2-devs@lists.launchpad.net >> Unsubscribe : https://launchpad.net/~dhis2-devs >> More help : https://help.launchpad.net/ListHelp >> >> -- *** Calle Hedberg 46D Alma Road, 7700 Rosebank, SOUTH AFRICA Tel/fax (home): +27-21-685-6472 Cell: +27-82-853-5352 Iridium SatPhone: +8816-315-19119 Email: calle.hedb...@gmail.com Skype: calle_hedberg *** ___ Mailing list: https://launchpad.net/~dhis2-devs Post to : dhis2-devs@lists.launchpad.net Unsubscribe : https://launchpad.net/~dhis2-devs More help : https://help.launchpad.net/ListHelp
Re: [Dhis2-devs] Import of Events is extremely slow - any plans to improve it?
Hi Calle, Thanks for your question. Just curious (I don't have answer). As we are planning to do this soon. Are you doing this through UI or API? If API, which endpoint enables this? Thanks Ime On Jun 28, 2017 19:15, "Calle Hedberg"wrote: > Hi > > I started importing around 700,000 events 2 days ago (about 5 mill > individual values) - and the import is still running, 48 hours later. > Import is slowing - it seemed to be importing around 7-8 values per second, > now down to 2-3 per second. It looks like that import might take around 200 > hours - which makes no sense from a usability point of view. Importing > OrgUnits used to be equally slow, but a revamp recently change a 10,000 > OrgUnit import from hours to something like 5 seconds. > > It seems obvious that event imports have been designed to import values > one by one, which is equivalent to asking people to fetch water using a > leaking tea-spoon. > > I've searched JIRA but cannot find any issue related to this, thus this > post: Are there any plans underway to make import of Event data usable for > larger record sets? > > Regards > Calle > > *** > > Calle Hedberg > > 46D Alma Road, 7700 Rosebank, SOUTH AFRICA > > Tel/fax (home): +27-21-685-6472 > > Cell: +27-82-853-5352 > > Iridium SatPhone: +8816-315-19119 > > Email: calle.hedb...@gmail.com > > Skype: calle_hedberg > > *** > > > ___ > Mailing list: https://launchpad.net/~dhis2-devs > Post to : dhis2-devs@lists.launchpad.net > Unsubscribe : https://launchpad.net/~dhis2-devs > More help : https://help.launchpad.net/ListHelp > > ___ Mailing list: https://launchpad.net/~dhis2-devs Post to : dhis2-devs@lists.launchpad.net Unsubscribe : https://launchpad.net/~dhis2-devs More help : https://help.launchpad.net/ListHelp
[Dhis2-devs] Import of Events is extremely slow - any plans to improve it?
Hi I started importing around 700,000 events 2 days ago (about 5 mill individual values) - and the import is still running, 48 hours later. Import is slowing - it seemed to be importing around 7-8 values per second, now down to 2-3 per second. It looks like that import might take around 200 hours - which makes no sense from a usability point of view. Importing OrgUnits used to be equally slow, but a revamp recently change a 10,000 OrgUnit import from hours to something like 5 seconds. It seems obvious that event imports have been designed to import values one by one, which is equivalent to asking people to fetch water using a leaking tea-spoon. I've searched JIRA but cannot find any issue related to this, thus this post: Are there any plans underway to make import of Event data usable for larger record sets? Regards Calle *** Calle Hedberg 46D Alma Road, 7700 Rosebank, SOUTH AFRICA Tel/fax (home): +27-21-685-6472 Cell: +27-82-853-5352 Iridium SatPhone: +8816-315-19119 Email: calle.hedb...@gmail.com Skype: calle_hedberg *** ___ Mailing list: https://launchpad.net/~dhis2-devs Post to : dhis2-devs@lists.launchpad.net Unsubscribe : https://launchpad.net/~dhis2-devs More help : https://help.launchpad.net/ListHelp