Good afternoon. My name is Camilo Freire, from Montevideo, Uruguay. I have a catalogue with around 280.000 bibliographic records in Koha 18.05.05. When I do some searches whose hit's number is about 14.000 records the time of search climbs from an average of 2 seconds per search to 40 to 80 seconds around. I've a virtual machine with 6GB of RAM, with a virtual Xeon of 2GHz. Koha is making the OPAC search calling the function getRecords which is in the lib/C4/Search.pm file. As I've seen this function first makes a call to the ZOOM::search function and then makes an asynchronous loop for getting the records (don't do that in ZOOM::search?) and facets, calling the _ZOOM_event_function. The Zebra search in ZOOM::search is quite fast, and takes no more than 0.3 seconds to perform the search. It seems that the delay is in the asynchronous facets in the _ZOOM_event_loop function. Is not clear to me what exactly the _ZOOM_event_loop is doing, apart from retrieving the facets, which are not necessary for me. Can I resolve in some way this problem, getting rid of the _ZOOM_event_loop or installing other version? As a first step I've enabled plack, which is a great help in speeding up the application but not in this case. Thank you very much in advance.
Greetings, I don't promise to be right, but here are a couple thoughts: 1) You can use Zebra Facets -- there's a koha-conf.xml line in there somewhere. 2) Also, do you have plack turned on? Plack increases response times considerably. GPML, Mark Tompsett
Actually, if you have Zebra facets enabled, turn them *off*. I believe that they're enabled by default, and they're dog slow. Zebra is *very* disk IO intensive. Make sure that its temporary directory is being written to the fastest drive you have. If you have the QueryFuzzy or QueryStemming system preferences enabled, be aware that they don't like single letter words such as "A" or "I" -- essentially, they end up trying to search for 'a*' or 'i*'; on a slow file system they can entirely time out and not return any results. Those system preferences *do* provide for easier searching -- disabling them means that keywords have to be typed exactly as they are in the record, so there's a trade-off between speed and user utility, but if the searches are too slow, it might be worth looking into. Finally, take a look at ElasticSearch. It isn't at feature-parity with Zebra in 18.05, but it might be worth a look. On Wed, Nov 7, 2018 at 6:23 PM Mark Tompsett <mtompset@hotmail.com> wrote:
Greetings,
I don't promise to be right, but here are a couple thoughts: 1) You can use Zebra Facets -- there's a koha-conf.xml line in there somewhere. 2) Also, do you have plack turned on? Plack increases response times considerably.
GPML, Mark Tompsett
_______________________________________________ Koha mailing list http://koha-community.org Koha@lists.katipo.co.nz https://lists.katipo.co.nz/mailman/listinfo/koha
A couple follow up questions for you, Barton. Are Zebra facets a general system preference? I see a few facets-related preferences, but nothing specifically tied to Zebra. We're still using Zebra in our system. Can you clarify what you mean by ElasticSearch not yet being at feature-parity in 18.05? We just upgraded to 18.05. Thoughts on staying w/ Zebra vs. switching over to ES? Many thanks in advance, Cab Vinton Plaistow Public Library Plaistow, NH On Thu, Nov 8, 2018 at 6:48 AM Barton Chittenden <barton@bywatersolutions.com> wrote:
Actually, if you have Zebra facets enabled, turn them *off*. I believe that they're enabled by default, and they're dog slow.
Zebra is *very* disk IO intensive. Make sure that its temporary directory is being written to the fastest drive you have.
If you have the QueryFuzzy or QueryStemming system preferences enabled, be aware that they don't like single letter words such as "A" or "I" -- essentially, they end up trying to search for 'a*' or 'i*'; on a slow file system they can entirely time out and not return any results. Those system preferences *do* provide for easier searching -- disabling them means that keywords have to be typed exactly as they are in the record, so there's a trade-off between speed and user utility, but if the searches are too slow, it might be worth looking into.
Finally, take a look at ElasticSearch. It isn't at feature-parity with Zebra in 18.05, but it might be worth a look.
On Wed, Nov 7, 2018 at 6:23 PM Mark Tompsett <mtompset@hotmail.com> wrote:
Greetings,
I don't promise to be right, but here are a couple thoughts: 1) You can use Zebra Facets -- there's a koha-conf.xml line in there somewhere. 2) Also, do you have plack turned on? Plack increases response times considerably.
GPML, Mark Tompsett
_______________________________________________ Koha mailing list http://koha-community.org Koha@lists.katipo.co.nz https://lists.katipo.co.nz/mailman/listinfo/koha
_______________________________________________ Koha mailing list http://koha-community.org Koha@lists.katipo.co.nz https://lists.katipo.co.nz/mailman/listinfo/koha
With respect to my last post on problems in OPAC Search in Koha, I've found that the problem is in the method _get_facets_from_zebra called from the method GetFacets of the class Search.pm. The asyncronous programming has nothing to do with the problem. Perhaps someone have had a similar problem or can help me about this. Thanks in advance Camilo Freire ---------- Forwarded message --------- From: Camilo Freire <camilofreire@gmail.com> Date: mié., 7 nov. 2018 a las 20:09 Subject: OPAC Search too slow in Koha 18.05.05 To: <koha@lists.katipo.co.nz> Good afternoon. My name is Camilo Freire, from Montevideo, Uruguay. I have a catalogue with around 280.000 bibliographic records in Koha 18.05.05. When I do some searches whose hit's number is about 14.000 records the time of search climbs from an average of 2 seconds per search to 40 to 80 seconds around. I've a virtual machine with 6GB of RAM, with a virtual Xeon of 2GHz. Koha is making the OPAC search calling the function getRecords which is in the lib/C4/Search.pm file. As I've seen this function first makes a call to the ZOOM::search function and then makes an asynchronous loop for getting the records (don't do that in ZOOM::search?) and facets, calling the _ZOOM_event_function. The Zebra search in ZOOM::search is quite fast, and takes no more than 0.3 seconds to perform the search. It seems that the delay is in the asynchronous facets in the _ZOOM_event_loop function. Is not clear to me what exactly the _ZOOM_event_loop is doing, apart from retrieving the facets, which are not necessary for me. Can I resolve in some way this problem, getting rid of the _ZOOM_event_loop or installing other version? As a first step I've enabled plack, which is a great help in speeding up the application but not in this case. Thank you very much in advance.
Hola Camilo, Take a look at the value of use_zebra_facets in your koha-conf.xml file. If enabled, disable it (see bug 13665). You may also want to play with the value of maxRecordsForFacets. Saludos, Jonathan https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=13665 On Thu, 8 Nov 2018 at 19:35 Camilo Freire <camilofreire@gmail.com> wrote:
With respect to my last post on problems in OPAC Search in Koha, I've found that the problem is in the method _get_facets_from_zebra called from the method GetFacets of the class Search.pm. The asyncronous programming has nothing to do with the problem.
Perhaps someone have had a similar problem or can help me about this.
Thanks in advance Camilo Freire
---------- Forwarded message --------- From: Camilo Freire <camilofreire@gmail.com> Date: mié., 7 nov. 2018 a las 20:09 Subject: OPAC Search too slow in Koha 18.05.05 To: <koha@lists.katipo.co.nz>
Good afternoon. My name is Camilo Freire, from Montevideo, Uruguay.
I have a catalogue with around 280.000 bibliographic records in Koha 18.05.05. When I do some searches whose hit's number is about 14.000 records the time of search climbs from an average of 2 seconds per search to 40 to 80 seconds around. I've a virtual machine with 6GB of RAM, with a virtual Xeon of 2GHz.
Koha is making the OPAC search calling the function getRecords which is in the lib/C4/Search.pm file. As I've seen this function first makes a call to the ZOOM::search function and then makes an asynchronous loop for getting the records (don't do that in ZOOM::search?) and facets, calling the _ZOOM_event_function. The Zebra search in ZOOM::search is quite fast, and takes no more than 0.3 seconds to perform the search. It seems that the delay is in the asynchronous facets in the _ZOOM_event_loop function. Is not clear to me what exactly the _ZOOM_event_loop is doing, apart from retrieving the facets, which are not necessary for me.
Can I resolve in some way this problem, getting rid of the _ZOOM_event_loop or installing other version? As a first step I've enabled plack, which is a great help in speeding up the application but not in this case.
Thank you very much in advance. _______________________________________________ Koha mailing list http://koha-community.org Koha@lists.katipo.co.nz https://lists.katipo.co.nz/mailman/listinfo/koha
participants (5)
-
Barton Chittenden -
Cab Vinton -
Camilo Freire -
Jonathan Druart -
Mark Tompsett