Re: q and logical operators.
Thanks for the heads up. On Fri, Sep 12, 2014 at 5:48 PM, Erick Erickson erickerick...@gmail.com wrote: John: Glad it worked. Bit a little careful with large slops. As the slop increases, you approach the same result set as vis AND dis AND dur so choosing the appropriate slop is something of a balancing act Best, Erick On Fri, Sep 12, 2014 at 2:10 AM, John Nielsen j...@mcb.dk wrote: I didn't know about sloppy queries. This is great stuff! I solved it with a qs=100. Thank you for the help. On Thu, Sep 11, 2014 at 11:36 PM, Erick Erickson erickerick...@gmail.com wrote: just skimmed, but: bq: I would get a hit for vis dis dur, but vis dur dis no longer returns anything. This is not an option for me Would slop help here? i.e. vis dur dis~3 or some such? Best Erick On Thu, Sep 11, 2014 at 4:34 AM, John Nielsen j...@mcb.dk wrote: q and logical operators. Hi all, I have a strange problem which seems to stomp my google-fu skills. We have a webshop which has a solr based search mechanism which allows customers to search for products based on a range of different fields, including item numbers. I recently added a feature which allows users who are logged in to search for custom item numbers which are associated with that user. What this means in practical terms is that when a user logs in, the solr search query has to look in one extra field compared to when the user is not logged in. The standard non-logged in search query looks like this (I only included the relevant first part of the query.): http:// secret/solr/11731_Danish/search?defType=edismaxq=Visitkort+display+Durable+4+rum+til+240+kort When doing the same search while logged in, the query looks like this: http:// secret/solr/11731_Danish/search?defType=edismaxq=Visitkort+display+Durable+4+rum+til+240+kort+OR+customer_5266762_product_number_string:Visitkort+display+Durable+4+rum+til+240+kort Here I add an extra field, customer_5266762_product_number_string (5266762 being the logged in users internal ID), basically including the same search tearm two times. The above examples work beautifully when searching for a specific item number stored in the customer_5266762_product_number_string. The problem is that when a user is logged in and want to do regular searches, the system begins to break down. In the specific example above, I expect to get a single hit for a product with the title Visitkort display Durable 4 rum til 240 kort. It works as expected with the first non-logged-in example. The second logged-in example returns over 7000 hits. I would expect it to return just one hit since there is nothing relevant in the customer_5266762_product_number_string for this query. Now, the following is where my brain begins to melt down. I discovered that if you put the search text in quotation marks, it will work as expected, but doing so breaks another loved feature we have: If i want a hit on the product named Visitkort display Durable 4 rum til 240 kort, I could do a search for vis dis dur, and it would show up. I could also get a hit if i write vis dur dis, changing the orden of the words. If i put the search query in quotation marks, I break that capability. I would get a hit for vis dis dur, but vis dur dis no longer returns anything. This is not an option for me. It is entirely posible that there is a better way of implementing this and fortunately, a rewrite is possible at this time. If my basic approach is correct and I just don't understand how to construct my query correctly, an RTFM pointer will be most welcome! -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: q and logical operators.
I didn't know about sloppy queries. This is great stuff! I solved it with a qs=100. Thank you for the help. On Thu, Sep 11, 2014 at 11:36 PM, Erick Erickson erickerick...@gmail.com wrote: just skimmed, but: bq: I would get a hit for vis dis dur, but vis dur dis no longer returns anything. This is not an option for me Would slop help here? i.e. vis dur dis~3 or some such? Best Erick On Thu, Sep 11, 2014 at 4:34 AM, John Nielsen j...@mcb.dk wrote: q and logical operators. Hi all, I have a strange problem which seems to stomp my google-fu skills. We have a webshop which has a solr based search mechanism which allows customers to search for products based on a range of different fields, including item numbers. I recently added a feature which allows users who are logged in to search for custom item numbers which are associated with that user. What this means in practical terms is that when a user logs in, the solr search query has to look in one extra field compared to when the user is not logged in. The standard non-logged in search query looks like this (I only included the relevant first part of the query.): http:// secret/solr/11731_Danish/search?defType=edismaxq=Visitkort+display+Durable+4+rum+til+240+kort When doing the same search while logged in, the query looks like this: http:// secret/solr/11731_Danish/search?defType=edismaxq=Visitkort+display+Durable+4+rum+til+240+kort+OR+customer_5266762_product_number_string:Visitkort+display+Durable+4+rum+til+240+kort Here I add an extra field, customer_5266762_product_number_string (5266762 being the logged in users internal ID), basically including the same search tearm two times. The above examples work beautifully when searching for a specific item number stored in the customer_5266762_product_number_string. The problem is that when a user is logged in and want to do regular searches, the system begins to break down. In the specific example above, I expect to get a single hit for a product with the title Visitkort display Durable 4 rum til 240 kort. It works as expected with the first non-logged-in example. The second logged-in example returns over 7000 hits. I would expect it to return just one hit since there is nothing relevant in the customer_5266762_product_number_string for this query. Now, the following is where my brain begins to melt down. I discovered that if you put the search text in quotation marks, it will work as expected, but doing so breaks another loved feature we have: If i want a hit on the product named Visitkort display Durable 4 rum til 240 kort, I could do a search for vis dis dur, and it would show up. I could also get a hit if i write vis dur dis, changing the orden of the words. If i put the search query in quotation marks, I break that capability. I would get a hit for vis dis dur, but vis dur dis no longer returns anything. This is not an option for me. It is entirely posible that there is a better way of implementing this and fortunately, a rewrite is possible at this time. If my basic approach is correct and I just don't understand how to construct my query correctly, an RTFM pointer will be most welcome! -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
q and logical operators.
q and logical operators. Hi all, I have a strange problem which seems to stomp my google-fu skills. We have a webshop which has a solr based search mechanism which allows customers to search for products based on a range of different fields, including item numbers. I recently added a feature which allows users who are logged in to search for custom item numbers which are associated with that user. What this means in practical terms is that when a user logs in, the solr search query has to look in one extra field compared to when the user is not logged in. The standard non-logged in search query looks like this (I only included the relevant first part of the query.): http:// secret/solr/11731_Danish/search?defType=edismaxq=Visitkort+display+Durable+4+rum+til+240+kort When doing the same search while logged in, the query looks like this: http:// secret/solr/11731_Danish/search?defType=edismaxq=Visitkort+display+Durable+4+rum+til+240+kort+OR+customer_5266762_product_number_string:Visitkort+display+Durable+4+rum+til+240+kort Here I add an extra field, customer_5266762_product_number_string (5266762 being the logged in users internal ID), basically including the same search tearm two times. The above examples work beautifully when searching for a specific item number stored in the customer_5266762_product_number_string. The problem is that when a user is logged in and want to do regular searches, the system begins to break down. In the specific example above, I expect to get a single hit for a product with the title Visitkort display Durable 4 rum til 240 kort. It works as expected with the first non-logged-in example. The second logged-in example returns over 7000 hits. I would expect it to return just one hit since there is nothing relevant in the customer_5266762_product_number_string for this query. Now, the following is where my brain begins to melt down. I discovered that if you put the search text in quotation marks, it will work as expected, but doing so breaks another loved feature we have: If i want a hit on the product named Visitkort display Durable 4 rum til 240 kort, I could do a search for vis dis dur, and it would show up. I could also get a hit if i write vis dur dis, changing the orden of the words. If i put the search query in quotation marks, I break that capability. I would get a hit for vis dis dur, but vis dur dis no longer returns anything. This is not an option for me. It is entirely posible that there is a better way of implementing this and fortunately, a rewrite is possible at this time. If my basic approach is correct and I just don't understand how to construct my query correctly, an RTFM pointer will be most welcome! -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Strange relevance scoring
Hi, We are seeing a strange phenomenon with our Solr setup which I have been unable to answer. My Google-fu is clearly not up to the task, so I am trying here. It appears that if i do a freetext search for a single word, say modellering on a text field, the scoring is massively boosted if the first word of the text field is a hit. For instance if there is only one occurrence of the word modellering in the text field and that occurrence is the first word of the text, then that document gets a higher relevancy than if the word modelling occurs 5 times in the text and the first word of the text is any other word. Is this normal behavior? Is special attention paid to the first word in a text field? I would think that the latter case would get the highest score. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Strange relevance scoring
Interesting. Most of the text fields are single word fields or close to it, but on some of the documents, long text appears. How long does a text need to be before hitting length normalization? On Tue, Apr 8, 2014 at 11:36 AM, Ahmet Arslan iori...@yahoo.com wrote: Hi Nielsen, There is no special attention paid to first word. You are probably hitting length normalisation. Lucene/Solr punishes long documents, favours short documents. (5 times appearing one) longer? On Tuesday, April 8, 2014 12:03 PM, John Nielsen j...@mcb.dk wrote: Hi, We are seeing a strange phenomenon with our Solr setup which I have been unable to answer. My Google-fu is clearly not up to the task, so I am trying here. It appears that if i do a freetext search for a single word, say modellering on a text field, the scoring is massively boosted if the first word of the text field is a hit. For instance if there is only one occurrence of the word modellering in the text field and that occurrence is the first word of the text, then that document gets a higher relevancy than if the word modelling occurs 5 times in the text and the first word of the text is any other word. Is this normal behavior? Is special attention paid to the first word in a text field? I would think that the latter case would get the highest score. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Strange relevance scoring
Hi, I couldn't find any occurrence of SpanFirstQuery in either the schema.xml or solrconfig.xml files. This is the query i used with debug=results. http://pastebin.com/bWzUkjKz And here is the answer. http://pastebin.com/nCXFcuky I am not sure what I am supposed to be looking for. On Tue, Apr 8, 2014 at 11:34 AM, Markus Jelsma markus.jel...@openindex.iowrote: Hi - the thing you describe is possible when your set up uses SpanFirstQuery. But to be sure what's going on you should post the debug output. -Original message- From:John Nielsen j...@mcb.dk Sent: Tuesday 8th April 2014 11:03 To: solr-user@lucene.apache.org Subject: Strange relevance scoring Hi, We are seeing a strange phenomenon with our Solr setup which I have been unable to answer. My Google-fu is clearly not up to the task, so I am trying here. It appears that if i do a freetext search for a single word, say modellering on a text field, the scoring is massively boosted if the first word of the text field is a hit. For instance if there is only one occurrence of the word modellering in the text field and that occurrence is the first word of the text, then that document gets a higher relevancy than if the word modelling occurs 5 times in the text and the first word of the text is any other word. Is this normal behavior? Is special attention paid to the first word in a text field? I would think that the latter case would get the highest score. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Informal poll on running Solr 4 on Java 7 with G1GC
We used to use G1, but recently went back to CMS. G1 gave us too long stop-the-world events. CMS uses more ressources for the same work, but it is more predictable and we get better worst-case performance out of it. Med venlig hilsen / Best regards John Nielsen Programmer MCB A/S Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On 20-06-2013 00:18, Timothy Potter wrote: I'm sure there's some site to do this but wanted to get a feel for who's running Solr 4 on Java 7 with G1 gc enabled? Cheers, Tim
Searching for cache stats
Hi, I am looking for an automated way of getting cache stats from Solr. Specificly what I am looking for are the cumulative evictions for each cache type for each core: http://screencast.com/t/IrD0VItfVduk An example of how I would like to be able to query the cache information is basically something like when I get core information like this: http://URL:8000/solr/admin/cores Does anything similar exist which will allow me to get the cache information? -- Med venlig hilsen / Best regards John Nielsen Programmer MCB A/S Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Solr using a ridiculous amount of memory
Sorry for not getting back to the list sooner. It seems like I finally solved the memory problems by following Toke's instruction of splitting the cores up into smaller chunks. After some major refactoring, our 15 cores have now turned into ~500 cores and our memory consumption has dropped dramaticly. Running 200 webshops now actually uses less memory as our 24 test shops did before. Thank you to everyone who helped, and especially to Toke. I looked at the wiki, but could not find any reference to this unintuitive way of using memory. Did I miss it somewhere? On Fri, Apr 19, 2013 at 1:30 PM, Erick Erickson erickerick...@gmail.comwrote: Hmmm. There has been quite a bit of work lately to support a couple of things that might be of interest (4.3, which Simon cut today, probably available to all mid next week at the latest). Basically, you can choose to pre-define all the cores in solr.xml (so-called old style) _or_ use the new-style solr.xml which uses auto-discover mode to walk the indicated directory and find all the cores (indicated by the presence of a 'core.properties' file). Don't know if this would make your particular case easier, and I should warn you that this is relatively new code (although there are some reasonable unit tests). You also have the option to only load the cores when they are referenced, and only keep N cores open at a time (loadOnStartup and transient properties). See: http://wiki.apache.org/solr/CoreAdmin#Configuration and http://wiki.apache.org/solr/Solr.xml%204.3%20and%20beyond Note, the docs are somewhat sketchy, so if you try to go down this route let us know anything that should be improved (or you can be added to the list of wiki page contributors and help out!) Best Erick On Thu, Apr 18, 2013 at 8:31 AM, John Nielsen j...@mcb.dk wrote: You are missing an essential part: Both the facet and the sort structures needs to hold one reference for each document _in_the_full_index_, even when the document does not have any values in the fields. Wow, thank you for this awesome explanation! This is where the penny dropped for me. I will definetely move to a multi-core setup. It will take some time and a lot of re-coding. As soon as I know the result, I will let you know! -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: SolrCloud loadbalancing, replication, and failover
jobs and have similar requirements. I would imagine that Google gets incredible response time because they have incredible amounts of RAM at their disposal that keep the important bits of their index instantly available. They have thousands of servers in each data center. I once got a look at the extent of Google's hardware in one data center - it was HUGE. I couldn't get in to examine things closely, they keep that stuff very locked down. Thanks, Shawn -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Solr using a ridiculous amount of memory
That was strange. As you are using a multi-valued field with the new setup, they should appear there. Yes, the new field we use for faceting is a multi valued field. Can you find the facet fields in any of the other caches? Yes, here it is, in the field cache: http://screencast.com/t/mAwEnA21yL I hope you are not calling the facets with facet.method=enum? Could you paste a typical facet-enabled search request? Here is a typical example (I added newlines for readability): http://172.22.51.111:8000/solr/default1_Danish/search ?defType=edismax q=*%3a* facet.field=%7b!ex%3dtagitemvariantoptions_int_mv_7+key%3ditemvariantoptions_int_mv_7%7ditemvariantoptions_int_mv facet.field=%7b!ex%3dtagitemvariantoptions_int_mv_9+key%3ditemvariantoptions_int_mv_9%7ditemvariantoptions_int_mv facet.field=%7b!ex%3dtagitemvariantoptions_int_mv_8+key%3ditemvariantoptions_int_mv_8%7ditemvariantoptions_int_mv facet.field=%7b!ex%3dtagitemvariantoptions_int_mv_2+key%3ditemvariantoptions_int_mv_2%7ditemvariantoptions_int_mv fq=site_guid%3a(10217) fq=item_type%3a(PRODUCT) fq=language_guid%3a(1) fq=item_group_1522_combination%3a(*) fq=is_searchable%3a(True) sort=item_group_1522_name_int+asc, variant_of_item_guid+asc querytype=Technical fl=feed_item_serialized facet=true group=true group.facet=true group.ngroups=true group.field=groupby_variant_of_item_guid group.sort=name+asc rows=0 Are you warming all the sort- and facet-fields? I'm sorry, I don't know. I have the field value cache commented out in my config, so... Whatever is default? Removing the custom sort fields is unfortunately quite a bit more difficult than my other facet modification. The problem is that each item can have several sort orders. The sort order to use is defined by a group number which is known ahead of time. The group number is included in the sort order field name. To solve it in the same way i solved the facet problem, I would need to be able to sort on a multi-valued field, and unless I'm wrong, I don't think that it's possible. I am quite stomped on how to fix this. On Wed, Apr 17, 2013 at 3:06 PM, Toke Eskildsen t...@statsbiblioteket.dkwrote: John Nielsen [j...@mcb.dk]: I never seriously looked at my fieldValueCache. It never seemed to get used: http://screencast.com/t/YtKw7UQfU That was strange. As you are using a multi-valued field with the new setup, they should appear there. Can you find the facet fields in any of the other caches? ...I hope you are not calling the facets with facet.method=enum? Could you paste a typical facet-enabled search request? Yep. We still do a lot of sorting on dynamic field names, so the field cache has a lot of entries. (9.411 entries as we speak. This is considerably lower than before.). You mentioned in an earlier mail that faceting on a field shared between all facet queries would bring down the memory needed. Does the same thing go for sorting? More or less. Sorting stores the raw string representations (utf-8) in memory so the number of unique values has more to say than it does for faceting. Just as with faceting, a list of pointers from documents to values (1 value/document as we are sorting) is maintained, so the overhead is something like #documents*log2(#unique_terms*average_term_length) + #unique_terms*average_term_length (where average_term_length is in bits) Caveat: This is with the index-wide sorting structure. I am fairly confident that this is what Solr uses, but I have not looked at it lately so it is possible that some memory-saving segment-based trickery has been implemented. Does those 9411 entries duplicate data between them? Sorry, I do not know. SOLR- discusses the problems with the field cache and duplication of data, but I cannot infer if it is has been solved or not. I am not familiar with the stat breakdown of the fieldCache, but it _seems_ to me that there are 2 or 3 entries for each segment for each sort field. Guesstimating further, let's say you have 30 segments in your index. Going with the guesswork, that would bring the number of sort fields to 9411/3/30 ~= 100. Looks like you use a custom sort field for each client? Extrapolating from 1.4M documents and 180 clients, let's say that there are 1.4M/180/5 unique terms for each sort-field and that their average length is 10. We thus have 1.4M*log2(1500*10*8) + 1500*10*8 bit ~= 23MB per sort field or about 4GB for all the 180 fields. With this few unique values, the doc-value structure is by far the biggest, just as with facets. As opposed to the faceting structure, this is fairly close to the actual memory usage. Switching to a single sort field would reduce the memory usage from 4GB to about 55MB. I do commit a bit more often than i should. I get these in my log file from time to time: PERFORMANCE WARNING: Overlapping onDeckSearchers=2 So 1 active searcher and 2 warming searchers. Ignoring that one of the warming searchers is highly likely
Re: Solr using a ridiculous amount of memory
http://172.22.51.111:8000/solr/default1_Danish/search [...] fq=site_guid%3a(10217) This constraints to hits to a specific customer, right? Any search will only be in a single customer's data? Yes, thats right. No search from any given client ever returns anything from another client. [Toke: Are you warming all the sort- and facet-fields?] I'm sorry, I don't know. I have the field value cache commented out in my config, so... Whatever is default? (a bit shaky here) I would say not warming. You could check simply by starting solr and looking at the caches before you issue any searches. The field cache shows 0 entries at startup. On the running server, forcing a commit (and thus opening a new searcher) does not change the number of entries. The problem is that each item can have several sort orders. The sort order to use is defined by a group number which is known ahead of time. The group number is included in the sort order field name. To solve it in the same way i solved the facet problem, I would need to be able to sort on a multi-valued field, and unless I'm wrong, I don't think that it's possible. That is correct. Three suggestions off the bat: 1) Reduce the number of sort fields by mapping names. Count the maximum number of unique sort fields for any given customer. That will be the total number of sort fields in the index. For each group number for a customer, map that number to one of the index-wide sort fields. This only works if the maximum number of unique fields is low (let's say a single field takes 50MB, so 20 fields should be okay). I just checked our DB. Our worst case scenario client has over a thousand groups for sorting. Granted, it may be, probably is, an error with the data. It is an interesting idea though and I will look into this posibility. 3) Switch to a layout where each customer has a dedicated core. The basic overhead is a lot larger than for a shared index, but it would make your setup largely immune to the adverse effect of many documents coupled with many facet- and sort-fields. Now this is where my brain melts down. If I understand the fieldCache mechanism correctly (which i can see that I don't), the data used for faceting and sorting is saved in the fieldCache using a key comprised of the fields used for said faceting/sorting. That data only contains the data which is actually used for the operation. This is what the fq queries are for. So if i generate a core for each client, I would have a client specific fieldCache containing the data from that client. Wouldn't I just split up the same data into several cores? I'm afraid I don't understand how this would help. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Solr using a ridiculous amount of memory
You are missing an essential part: Both the facet and the sort structures needs to hold one reference for each document _in_the_full_index_, even when the document does not have any values in the fields. Wow, thank you for this awesome explanation! This is where the penny dropped for me. I will definetely move to a multi-core setup. It will take some time and a lot of re-coding. As soon as I know the result, I will let you know! -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Solr using a ridiculous amount of memory
I managed to get this done. The facet queries now facets on a multivalue field as opposed to the dynamic field names. Unfortunately it doesn't seem to have done much difference, if any at all. Some more information that might help: The JVM memory seem to be eaten up slowly. I dont think that there is one single query that causes the problem. My test case (dumping 180 clients on top of solr) takes hours before it causes an OOM. Often a full day. The memory usage wobbles up and down, so the GC is at least partially doing its job. It still works its way up to 100% eventually. When that happens it either OOM's or it stops the world and brings the memory consumption to 10-15 gigs. I did try to facet on all products across all clients (about 1.4 mil docs) and i could not make it OOM on a server with a 4 gig jvm. This was on a dedicated test server with my test being the only traffic. I am beginning to think that this may be related to traffic volume and not just on the type of query that I do. I tried to calculate the memory requirement example you gave me above based on the change that got rid of the dynamic fields. documents = ~1.400.000 references 11.200.000 (we facet on two multivalue fields with each 4 values on average, so 1.400.000 * 2 * 4 = 11.200.000 unique values = 1.132.344 (total number of variant options across all clients. This is what we facet on) 1.400.000 * log2(11.200.000) + 1.400.000 * log2(1132344) = ~14MB per field (we have 4 fields)? I must be calculating this wrong. On Mon, Apr 15, 2013 at 2:10 PM, John Nielsen j...@mcb.dk wrote: I did a search. I have no occurrence of UnInverted in the solr logs. Another explanation for the large amount of memory presents itself if you use a single index: If each of your clients facet on at least one fields specific to the client (client123_persons or something like that), then your memory usage goes through the roof. This is exactly how we facet right now! I will definetely rewrite the relevant parts of our product to test this out before moving further down the docValues path. I will let you know as soon as I know one way or the other. On Mon, Apr 15, 2013 at 1:38 PM, Toke Eskildsen t...@statsbiblioteket.dkwrote: On Mon, 2013-04-15 at 10:25 +0200, John Nielsen wrote: The FieldCache is the big culprit. We do a huge amount of faceting so it seems right. Yes, you wrote that earlier. The mystery is that the math does not check out with the description you have given us. Unfortunately I am super swamped at work so I have precious little time to work on this, which is what explains my silence. No problem, we've all been there. [Band aid: More memory] The extra memory helped a lot, but it still OOM with about 180 clients using it. You stated earlier that you has a solr cluster and your total(?) index size was 35GB, with each register being between 15k and 30k. I am using the quotes to signify that it is unclear what you mean. Is your cluster multiple machines (I'm guessing no), multiple Solr's, cores, shards or maybe just a single instance prepared for later distribution? Is a register a core, shard or a simply logical part (one client's data) of the index? If each client has their own core or shard, that would mean that each client uses more than 25GB/180 bytes ~= 142MB of heap to access 35GB/180 ~= 200MB of index. That sounds quite high and you would need a very heavy facet to reach that. If you could grep UnInverted from the Solr log file and paste the entries here, that would help to clarify things. Another explanation for the large amount of memory presents itself if you use a single index: If each of your clients facet on at least one fields specific to the client (client123_persons or something like that), then your memory usage goes through the roof. Assuming an index with 10M documents, each with 5 references to a modest 10K unique values in a facet field, the simplified formula #documents*log2(#references) + #references*log2(#unique_values) bit tells us that this takes at least 110MB with field cache based faceting. 180 clients @ 110MB ~= 20GB. As that is a theoretical low, we can at least double that. This fits neatly with your new heap of 64GB. If my guessing is correct, you can solve your memory problems very easily by sharing _all_ the facet fields between your clients. This should bring your memory usage down to a few GB. You are probably already restricting their searches to their own data by filtering, so this should not influence the returned facet values and counts, as compared to separate fields. This is very similar to the thread Facets with 5000 facet fields BTW. Today I finally managed to set up a test core so I can begin to play around with docValues. If you are using a single index with the individual-facet-fields for each client approach, the DocValues will also have scaling issues, as the amount of values (of which
Re: Solr using a ridiculous amount of memory
I am surprised about the lack of UnInverted from your logs as it is logged on INFO level. Nope, no trace of it. No mention either in Logging - Level from the admin interface. It should also be available from the admin interface under collection/Plugin / Stats/CACHE/fieldValueCache. I never seriously looked at my fieldValueCache. It never seemed to get used: http://screencast.com/t/YtKw7UQfU You stated that you were unable to make a 4GB JVM OOM when you just performed faceting (I guesstimate that it will also run fine with just ½GB or at least with 1GB, based on the numbers above) and you have observed that the field cache eats the memory. Yep. We still do a lot of sorting on dynamic field names, so the field cache has a lot of entries. (9.411 entries as we speak. This is considerably lower than before.). You mentioned in an earlier mail that faceting on a field shared between all facet queries would bring down the memory needed. Does the same thing go for sorting? Does those 9411 entries duplicate data between them? If this is where all the memory is going, I have a lot of coding to do. Guessing wildly: Do you issue a high frequency small updates with frequent commits? If you pause the indexing, does memory use fall back to the single GB level I do commit a bit more often than i should. I get these in my log file from time to time: PERFORMANCE WARNING: Overlapping onDeckSearchers=2 The way I understand this is that two searchers are being warmed at the same time and that one will be discarded when it finishes its auto warming procedure. If the math above is correct, I would need tens of searchers auto warming in parallel to cause my problem. If I misunderstand how this works, do let me know. My indexer has a cleanup routine that deletes replay logs and other things when it has nothing to do. This includes running a commit on the solr server to make sure nothing is ever in a state where something is not written to disk anywhere. In theory it can commit once every 60 seconds, though i doubt that ever happenes. The less work the indexer has, the more often it commits. (yes i know, its on my todo list) Other than that, my autocommit settings look like this: autoCommit maxTime6/maxTime maxDocs6000/maxDocs openSearcher false/openSearcher /autoCommit The control panel says that the warm up time of the last searcher is 5574. Is that seconds or milliseconds? http://screencast.com/t/d9oIbGLCFQwl I would prefer to not turn off the indexer unless the numbers above suggests that I really should try this. Waiting for a full GC would take a long time. Unfortunately I don't know of a way to provoke a full GC on command. On Wed, Apr 17, 2013 at 11:48 AM, Toke Eskildsen t...@statsbiblioteket.dkwrote: John Nielsen [j...@mcb.dk] wrote: I managed to get this done. The facet queries now facets on a multivalue field as opposed to the dynamic field names. Unfortunately it doesn't seem to have done much difference, if any at all. I am sorry to hear that. documents = ~1.400.000 references 11.200.000 (we facet on two multivalue fields with each 4 values on average, so 1.400.000 * 2 * 4 = 11.200.000 unique values = 1.132.344 (total number of variant options across all clients. This is what we facet on) 1.400.000 * log2(11.200.000) + 1.400.000 * log2(1132344) = ~14MB per field (we have 4 fields)? I must be calculating this wrong. No, that sounds about right. In reality you need to multiply with 3 or 4, so let's round to 50MB/field: 1.4M documents with 2 fields with 5M references/field each is not very much and should not take a lot of memory. In comparison, we facet on 12M documents with 166M references and do some other stuff (in Lucene with a different faceting implementation, but at this level it is equivalent to Solr's in terms of memory). Our heap is 3GB. I am surprised about the lack of UnInverted from your logs as it is logged on INFO level. It should also be available from the admin interface under collection/Plugin / Stats/CACHE/fieldValueCache. But I am guessing you got your numbers from that and that the list only contains the few facets you mentioned previously? It might be wise to sanity check by summing the memSizes though; they ought to take up far below 1GB. From your description, your index is small and your faceting requirements modest. A SSD-equipped laptop should be adequate as server. So we are back to math does not check out. You stated that you were unable to make a 4GB JVM OOM when you just performed faceting (I guesstimate that it will also run fine with just ½GB or at least with 1GB, based on the numbers above) and you have observed that the field cache eats the memory. This does indicate that the old caches are somehow not freed when the index is updated. That is strange as Solr should take care of that automatically. Guessing wildly: Do you issue a high frequency small updates with frequent commits? If you pause
Re: Solr using a ridiculous amount of memory
Yes and no, The FieldCache is the big culprit. We do a huge amount of faceting so it seems right. Unfortunately I am super swamped at work so I have precious little time to work on this, which is what explains my silence. Out of desperation, I added another 32G of memory to each server and increased the JVM size to 64G from 25G. The servers are running with 96G memory right now (this is the max amount supported by the hardware) which leaves solr somewhat starved for memory. I am aware of the performance implications of doing this but I have little choice. The extra memory helped a lot, but it still OOM with about 180 clients using it. Unfortunately I need to support at least double that. After upgrading the RAM, I ran for almost two weeks with the same workload that used to OOM a couple of times a day, so it doesn't look like a leak. Today I finally managed to set up a test core so I can begin to play around with docValues. I actually have a couple of questions regarding docValues: 1) If I facet on multible fields and only some of those fields are using docValues, will I still get the memory saving benefit of docValues? (one of the facet fields use null values and will require a lot of work in our product to fix) 2) If i just use docValues on one small core with very limited traffic at first for testing purposes, how can I test that it is actually using the disk for caching? I really appreciate all the help I have received on this list so far. I do feel confident that I will be able to solve this issue eventually. On Mon, Apr 15, 2013 at 9:00 AM, Toke Eskildsen t...@statsbiblioteket.dkwrote: On Sun, 2013-03-24 at 09:19 +0100, John Nielsen wrote: Our memory requirements are running amok. We have less than a quarter of our customers running now and even though we have allocated 25GB to the JVM already, we are still seeing daily OOM crashes. Out of curiosity: Did you manage to pinpoint the memory eater in your setup? - Toke Eskildsen -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Solr using a ridiculous amount of memory
I did a search. I have no occurrence of UnInverted in the solr logs. Another explanation for the large amount of memory presents itself if you use a single index: If each of your clients facet on at least one fields specific to the client (client123_persons or something like that), then your memory usage goes through the roof. This is exactly how we facet right now! I will definetely rewrite the relevant parts of our product to test this out before moving further down the docValues path. I will let you know as soon as I know one way or the other. On Mon, Apr 15, 2013 at 1:38 PM, Toke Eskildsen t...@statsbiblioteket.dkwrote: On Mon, 2013-04-15 at 10:25 +0200, John Nielsen wrote: The FieldCache is the big culprit. We do a huge amount of faceting so it seems right. Yes, you wrote that earlier. The mystery is that the math does not check out with the description you have given us. Unfortunately I am super swamped at work so I have precious little time to work on this, which is what explains my silence. No problem, we've all been there. [Band aid: More memory] The extra memory helped a lot, but it still OOM with about 180 clients using it. You stated earlier that you has a solr cluster and your total(?) index size was 35GB, with each register being between 15k and 30k. I am using the quotes to signify that it is unclear what you mean. Is your cluster multiple machines (I'm guessing no), multiple Solr's, cores, shards or maybe just a single instance prepared for later distribution? Is a register a core, shard or a simply logical part (one client's data) of the index? If each client has their own core or shard, that would mean that each client uses more than 25GB/180 bytes ~= 142MB of heap to access 35GB/180 ~= 200MB of index. That sounds quite high and you would need a very heavy facet to reach that. If you could grep UnInverted from the Solr log file and paste the entries here, that would help to clarify things. Another explanation for the large amount of memory presents itself if you use a single index: If each of your clients facet on at least one fields specific to the client (client123_persons or something like that), then your memory usage goes through the roof. Assuming an index with 10M documents, each with 5 references to a modest 10K unique values in a facet field, the simplified formula #documents*log2(#references) + #references*log2(#unique_values) bit tells us that this takes at least 110MB with field cache based faceting. 180 clients @ 110MB ~= 20GB. As that is a theoretical low, we can at least double that. This fits neatly with your new heap of 64GB. If my guessing is correct, you can solve your memory problems very easily by sharing _all_ the facet fields between your clients. This should bring your memory usage down to a few GB. You are probably already restricting their searches to their own data by filtering, so this should not influence the returned facet values and counts, as compared to separate fields. This is very similar to the thread Facets with 5000 facet fields BTW. Today I finally managed to set up a test core so I can begin to play around with docValues. If you are using a single index with the individual-facet-fields for each client approach, the DocValues will also have scaling issues, as the amount of values (of which the majority will be null) will be #clients*#documents*#facet_fields This means that the adding a new client will be progressively more expensive. On the other hand, if you use a lot of small shards, DocValues should work for you. Regards, Toke Eskildsen -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Solr using a ridiculous amount of memory
I apologize for the slow reply. Today has been killer. I will reply to everyone as soon as I get the time. I am having difficulties understanding how docValues work. Should I only add docValues to the fields that I actually use for sorting and faceting or on all fields? Will the docValues magic apply to the fields i activate docValues on or on the entire document when sorting/faceting on a field that has docValues activated? I'm not even sure which question to ask. I am struggling to understand this on a conceptual level. On Sun, Mar 24, 2013 at 7:11 PM, Robert Muir rcm...@gmail.com wrote: On Sun, Mar 24, 2013 at 4:19 AM, John Nielsen j...@mcb.dk wrote: Schema with DocValues attempt at solving problem: http://pastebin.com/Ne23NnW4 Config: http://pastebin.com/x1qykyXW This schema isn't using docvalues, due to a typo in your config. it should not be DocValues=true but docValues=true. Are you not getting an error? Solr needs to throw exception if you provide invalid attributes to the field. Nothing is more frustrating than having a typo or something in your configuration and solr just ignores this, reports no error, and doesnt work the way you want. I'll look into this (I already intend to add these checks to analysis factories for the same reason). Separately, if you really want the terms data and so on to remain on disk, it is not enough to just enable docvalues for the field. The default implementation uses the heap. So if you want that, you need to set docValuesFormat=Disk on the fieldtype. This will keep the majority of the data on disk, and only some key datastructures in heap memory. This might have significant performance impact depending upon what you are doing so you need to test that. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Solr using a ridiculous amount of memory
Hello all, We are running a solr cluster which is now running solr-4.2. The index is about 35GB on disk with each register between 15k and 30k. (This is simply the size of a full xml reply of one register. I'm not sure how to measure it otherwise.) Our memory requirements are running amok. We have less than a quarter of our customers running now and even though we have allocated 25GB to the JVM already, we are still seeing daily OOM crashes. We used to just allocate more memory to the JVM, but with the way solr is scaling, we would need well over 100GB of memory on each node to finish the project, and thats just not going to happen. I need to lower the memory requirements somehow. I can see from the memory dumps we've done that the field cache is by far the biggest sinner. Of special interest to me is the recent introduction of DocValues which supposedly mitigates this issue by using memory outside the JVM. I just can't, because of lack of documentation, seem to make it work. We do a lot of facetting. One client facets on about 50.000 docs of approx 30k each on 5 fields. I understand that this is VERY memory intensive. Schema with DocValues attempt at solving problem: http://pastebin.com/Ne23NnW4 Config: http://pastebin.com/x1qykyXW The cache is pretty well tuned. Any lower and i get evictions. Come hell or high water, my JVM memory requirements must come down. Simply moving some memory load outside of the JVM would be awesome! Making it not use the field cache for anything would also (probably) work for me. I thought about killing off my other caches, but from the dumps, they just don't seem to use that much memory. I am at my wits end. Any help would be sorely appreciated. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: SOLR - Documents with large number of fields ~ 450
with the on disk option. Could you elaborate on that? Den 22/03/2013 05.25 skrev Mark Miller markrmil...@gmail.com: You might try using docvalues with the on disk option and try and let the OS manage all the memory needed for all the faceting/sorting. This would require Solr 4.2. - Mark On Mar 21, 2013, at 2:56 AM, kobe.free.wo...@gmail.com wrote: Hello All, Scenario: My data model consist of approx. 450 fields with different types of data. We want to include each field for indexing as a result it will create a single SOLR document with *450 fields*. The total of number of records in the data set is *755K*. We will be using the features like faceting and sorting on approx. 50 fields. We are planning to use SOLR 4.1. Following is the hardware configuration of the web server that we plan to install SOLR on:- CPU: 2 x Dual Core (4 cores) | RAM: 12GB | Storage: 212 GB Questions : 1)What's the best approach when dealing with documents with large number of fields. What's the drawback of having a single document with a very large number of fields. Does SOLR support documents with large number of fields as in my case? 2)Will there be any performance issue if i define all of the 450 fields for indexing? Also if faceting is done on 50 fields with document having large number of fields and huge number of records? 3)The name of the fields in the data set are quiet lengthy around 60 characters. Will it be a problem defining fields with such a huge name in the schema file? Is there any best practice to be followed related to naming convention? Will big field names create problem during querying? Thanks! -- View this message in context: http://lucene.472066.n3.nabble.com/SOLR-Documents-with-large-number-of-fields-450-tp4049633.html Sent from the Solr - User mailing list archive at Nabble.com.
Re: Facets with 5000 facet fields
It looks like docvalues might solve a problem we have. (sorry for the thread jacking) I looked for info on it on the wiki, but could not find any. Is there any documentation done on it yet? On Wed, Mar 20, 2013 at 6:09 PM, Mark Miller markrmil...@gmail.com wrote: On Mar 20, 2013, at 11:29 AM, Chris Hostetter hossman_luc...@fucit.org wrote: Not true ... per segment FIeldCache support is available in solr faceting, you just have to specify facet.method=fcs (FieldCache per Segment) Also, if you use docvalues in 4.2, Robert tells me it is uses a new per seg faceting method that may have some better nrt characteristics than fcs. I have not played with it yet but hope to soon. - Mark -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: Data from deleted from Solr (Solr cloud)
Yeah, I ran into this issue myself with solr-4.0.0. To fix it, I had to compile my own version from the solr-4x branch. That is, I assume it's fixed as I have been unable to replicate it after the switch. I'm afraid you will have to reindex your data. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Wed, Dec 19, 2012 at 5:08 PM, shreejay shreej...@gmail.com wrote: Hi All, I have a solrlcoud instance with 3 shards. Each shard has 2 instance (2 servers each running a instance of solr) Lets say I had Instance1 and instance2 in shard1 … At some point, instance2 went down due to OOM (out of memory) . instance1 for some reason was not replicating the data properly and when it became the leader, it had only around 1% of the data that instance2 had. I restarted instance2, and hoped that instance1 will replicate from 2, but instead instanace2 replicated from instance1 . and ended up deleting the original index folder it had. There were around 2 million documents in that instance. Can any one of solrlcoud users give any hints if I can recover this data? --Shreejay -- View this message in context: http://lucene.472066.n3.nabble.com/Data-from-deleted-from-Solr-Solr-cloud-tp4028055.html Sent from the Solr - User mailing list archive at Nabble.com.
Re: Strange data-loss problem on one of our cores
I build a solr version from the solr-4x branch yesterday and so far am unable to replicate the problems i had before. I am cautiously optimistic that the problem has been resolved. If i run into any more problems, I'll let you all know. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 7:33 PM, Markus Jelsma markus.jel...@openindex.iowrote: Mark, no issue has been filed. That cluster runs a check out from round end of july/beginning of august. I'm in the process of including another cluster in the indexing and removal of documents besides the old production clusters. I'll start writing to that one tuesday orso. If i notice a discrepancy after some time i am sure to report it. I doubt i'll find it before 2013, if the problem is still there. -Original message- From:Mark Miller markrmil...@gmail.com Sent: Fri 14-Dec-2012 19:05 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores Have you filed a JIRA issue for this that I don't remember Markus? We need to make sure this is fixed. Any idea around when the trunk version came from? Before or after 4.0? - Mark On Dec 14, 2012, at 6:36 AM, Markus Jelsma markus.jel...@openindex.io wrote: We did not solve it but reindexing can remedy the problem. -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 12:31 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores How did you solve the problem? -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 12:04 PM, Markus Jelsma markus.jel...@openindex.iowrote: FYI, we observe the same issue, after some time (days, months) a cluster running an older trunk version has at least two shards where the leader and the replica do not contain the same number of records. No recovery is attempted, it seems it thinks everything is alright. Also, one core of one of the unsynced shards waits forever loading /replication?command=detailwt=json, other cores load it in a few ms. Both cores of another unsynced shard does not show this problem. -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 11:50 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores I did a manual commit, and we are still missing docs, so it doesn't look like the search race condition you mention. My boss wasn't happy when i mentioned that I wanted to try out unreleased code. Ill get him won over though and return with my findings. It will probably be some time next week. Thanks for your help. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Dec 13, 2012 at 4:10 PM, Mark Miller markrmil...@gmail.com wrote: Couple things to start: By default SolrCloud distributes updates a doc at a time. So if you have 1 shard, whatever node you index too, it will send updates to the other. Replication is only used for recovery, not distributing data. So for some reason, there is an IOException when it tries to forward. The other issue is not something that Ive seen reported. Can/did you try and do another hard commit to make sure you had the latest search open when checking the # of docs on each node? There was previously a race around commit that could cause some issues around expected visibility. If you are able to, you might try out a nightly build - 4.1 will be ready very soon and has numerous bug fixes for SolrCloud. - Mark On Dec 13, 2012, at 9:53 AM, John Nielsen j...@mcb.dk wrote: Hi all, We are seeing a strange problem on our 2-node solr4 cluster. This problem has resultet in data loss. We have two servers, varnish01 and varnish02. Zookeeper is running on varnish02, but in a separate jvm. We index directly to varnish02 and we read from varnish01. Data is thus replicated from varnish02 to varnish01. I found this in the varnish01 log: *INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=42 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr
Re: Strange data-loss problem on one of our cores
I did a manual commit, and we are still missing docs, so it doesn't look like the search race condition you mention. My boss wasn't happy when i mentioned that I wanted to try out unreleased code. Ill get him won over though and return with my findings. It will probably be some time next week. Thanks for your help. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Dec 13, 2012 at 4:10 PM, Mark Miller markrmil...@gmail.com wrote: Couple things to start: By default SolrCloud distributes updates a doc at a time. So if you have 1 shard, whatever node you index too, it will send updates to the other. Replication is only used for recovery, not distributing data. So for some reason, there is an IOException when it tries to forward. The other issue is not something that Ive seen reported. Can/did you try and do another hard commit to make sure you had the latest search open when checking the # of docs on each node? There was previously a race around commit that could cause some issues around expected visibility. If you are able to, you might try out a nightly build - 4.1 will be ready very soon and has numerous bug fixes for SolrCloud. - Mark On Dec 13, 2012, at 9:53 AM, John Nielsen j...@mcb.dk wrote: Hi all, We are seeing a strange problem on our 2-node solr4 cluster. This problem has resultet in data loss. We have two servers, varnish01 and varnish02. Zookeeper is running on varnish02, but in a separate jvm. We index directly to varnish02 and we read from varnish01. Data is thus replicated from varnish02 to varnish01. I found this in the varnish01 log: *INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=42 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=41 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=33 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=33 Dec 13, 2012 12:23:39 PM org.apache.solr.common.SolrException log SEVERE: shard update error StdNode: http://varnish02.lynero.net:8000/solr/default1_Norwegian/:org.apache.solr.client.solrj.SolrServerException : IOException occured when talking to server at: http://varnish02.lynero.net:8000/solr/default1_Norwegian at org.apache.solr.client.solrj.impl.HttpSolrServer.request(HttpSolrServer.java:413) at org.apache.solr.client.solrj.impl.HttpSolrServer.request(HttpSolrServer.java:181) at org.apache.solr.update.SolrCmdDistributor$1.call(SolrCmdDistributor.java:335) at org.apache.solr.update.SolrCmdDistributor$1.call(SolrCmdDistributor.java:309) at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:334) at java.util.concurrent.FutureTask.run(FutureTask.java:166) at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471) at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:334) at java.util.concurrent.FutureTask.run(FutureTask.java:166) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1110) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:603) at java.lang.Thread.run(Thread.java:636) Caused by: org.apache.http.NoHttpResponseException: The target server failed to respond at org.apache.http.impl.conn.DefaultResponseParser.parseHead(DefaultResponseParser.java:101) at org.apache.http.impl.io.AbstractMessageParser.parse(AbstractMessageParser.java:252) at org.apache.http.impl.AbstractHttpClientConnection.receiveResponseHeader(AbstractHttpClientConnection.java:282) at org.apache.http.impl.conn.DefaultClientConnection.receiveResponseHeader(DefaultClientConnection.java:247) at org.apache.http.impl.conn.AbstractClientConnAdapter.receiveResponseHeader(AbstractClientConnAdapter.java:216) at org.apache.http.protocol.HttpRequestExecutor.doReceiveResponse(HttpRequestExecutor.java:298) at org.apache.http.protocol.HttpRequestExecutor.execute(HttpRequestExecutor.java:125) at org.apache.http.impl.client.DefaultRequestDirector.tryExecute(DefaultRequestDirector.java:647
Re: Strange data-loss problem on one of our cores
How did you solve the problem? -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 12:04 PM, Markus Jelsma markus.jel...@openindex.iowrote: FYI, we observe the same issue, after some time (days, months) a cluster running an older trunk version has at least two shards where the leader and the replica do not contain the same number of records. No recovery is attempted, it seems it thinks everything is alright. Also, one core of one of the unsynced shards waits forever loading /replication?command=detailwt=json, other cores load it in a few ms. Both cores of another unsynced shard does not show this problem. -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 11:50 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores I did a manual commit, and we are still missing docs, so it doesn't look like the search race condition you mention. My boss wasn't happy when i mentioned that I wanted to try out unreleased code. Ill get him won over though and return with my findings. It will probably be some time next week. Thanks for your help. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Dec 13, 2012 at 4:10 PM, Mark Miller markrmil...@gmail.com wrote: Couple things to start: By default SolrCloud distributes updates a doc at a time. So if you have 1 shard, whatever node you index too, it will send updates to the other. Replication is only used for recovery, not distributing data. So for some reason, there is an IOException when it tries to forward. The other issue is not something that Ive seen reported. Can/did you try and do another hard commit to make sure you had the latest search open when checking the # of docs on each node? There was previously a race around commit that could cause some issues around expected visibility. If you are able to, you might try out a nightly build - 4.1 will be ready very soon and has numerous bug fixes for SolrCloud. - Mark On Dec 13, 2012, at 9:53 AM, John Nielsen j...@mcb.dk wrote: Hi all, We are seeing a strange problem on our 2-node solr4 cluster. This problem has resultet in data loss. We have two servers, varnish01 and varnish02. Zookeeper is running on varnish02, but in a separate jvm. We index directly to varnish02 and we read from varnish01. Data is thus replicated from varnish02 to varnish01. I found this in the varnish01 log: *INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=42 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=41 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=33 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=33 Dec 13, 2012 12:23:39 PM org.apache.solr.common.SolrException log SEVERE: shard update error StdNode: http://varnish02.lynero.net:8000/solr/default1_Norwegian/:org.apache.solr.client.solrj.SolrServerException : IOException occured when talking to server at: http://varnish02.lynero.net:8000/solr/default1_Norwegian at org.apache.solr.client.solrj.impl.HttpSolrServer.request(HttpSolrServer.java:413) at org.apache.solr.client.solrj.impl.HttpSolrServer.request(HttpSolrServer.java:181) at org.apache.solr.update.SolrCmdDistributor$1.call(SolrCmdDistributor.java:335) at org.apache.solr.update.SolrCmdDistributor$1.call(SolrCmdDistributor.java:309) at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:334) at java.util.concurrent.FutureTask.run(FutureTask.java:166) at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471) at java.util.concurrent.FutureTask$Sync.innerRun(FutureTask.java:334) at java.util.concurrent.FutureTask.run(FutureTask.java:166
Re: Strange data-loss problem on one of our cores
I'm building a simple tool which will help us monitor the solr cores for this problem. Basically it does a q=*:* on both servers on each cores and compares numFound of each result. Problem is that since this is a cloud setup, i can't be sure which server gets me the result. Is there a parameter I can add to the GET requests that will lock the request to a specific node in the cluster, treating the server receiving the request as a standalone server as opposed to a member of a cluster? I tried googeling it without luck. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 12:36 PM, Markus Jelsma markus.jel...@openindex.iowrote: We did not solve it but reindexing can remedy the problem. -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 12:31 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores How did you solve the problem? -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 12:04 PM, Markus Jelsma markus.jel...@openindex.iowrote: FYI, we observe the same issue, after some time (days, months) a cluster running an older trunk version has at least two shards where the leader and the replica do not contain the same number of records. No recovery is attempted, it seems it thinks everything is alright. Also, one core of one of the unsynced shards waits forever loading /replication?command=detailwt=json, other cores load it in a few ms. Both cores of another unsynced shard does not show this problem. -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 11:50 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores I did a manual commit, and we are still missing docs, so it doesn't look like the search race condition you mention. My boss wasn't happy when i mentioned that I wanted to try out unreleased code. Ill get him won over though and return with my findings. It will probably be some time next week. Thanks for your help. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Dec 13, 2012 at 4:10 PM, Mark Miller markrmil...@gmail.com wrote: Couple things to start: By default SolrCloud distributes updates a doc at a time. So if you have 1 shard, whatever node you index too, it will send updates to the other. Replication is only used for recovery, not distributing data. So for some reason, there is an IOException when it tries to forward. The other issue is not something that Ive seen reported. Can/did you try and do another hard commit to make sure you had the latest search open when checking the # of docs on each node? There was previously a race around commit that could cause some issues around expected visibility. If you are able to, you might try out a nightly build - 4.1 will be ready very soon and has numerous bug fixes for SolrCloud. - Mark On Dec 13, 2012, at 9:53 AM, John Nielsen j...@mcb.dk wrote: Hi all, We are seeing a strange problem on our 2-node solr4 cluster. This problem has resultet in data loss. We have two servers, varnish01 and varnish02. Zookeeper is running on varnish02, but in a separate jvm. We index directly to varnish02 and we read from varnish01. Data is thus replicated from varnish02 to varnish01. I found this in the varnish01 log: *INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=42 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=41 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr path=/update params={distrib.from= http://varnish02.lynero.net:8000/solr/default1_Norwegian/update.distrib=TOLEADERwt=javabinversion=2 } status=0 QTime=33 Dec 13, 2012 12:23:36 PM org.apache.solr.core.SolrCore execute INFO: [default1_Norwegian] webapp=/solr
Re: Strange data-loss problem on one of our cores
Awesome! http://host:port/solr/admin/cores is exactly what i needed! -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 1:21 PM, Markus Jelsma markus.jel...@openindex.iowrote: You must use the core's name and not use the collection name so you have to know which core is on which server. http://host:port/solr/corename/select You can use the cores handler to find out about the cores on the node: http://host:port/solr/admin/cores You can also use luke for this. It returns the same stats as in the interface: http://host:port/solr/corename/admin/luke -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 13:16 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores I'm building a simple tool which will help us monitor the solr cores for this problem. Basically it does a q=*:* on both servers on each cores and compares numFound of each result. Problem is that since this is a cloud setup, i can't be sure which server gets me the result. Is there a parameter I can add to the GET requests that will lock the request to a specific node in the cluster, treating the server receiving the request as a standalone server as opposed to a member of a cluster? I tried googeling it without luck. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 12:36 PM, Markus Jelsma markus.jel...@openindex.iowrote: We did not solve it but reindexing can remedy the problem. -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 12:31 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores How did you solve the problem? -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Dec 14, 2012 at 12:04 PM, Markus Jelsma markus.jel...@openindex.iowrote: FYI, we observe the same issue, after some time (days, months) a cluster running an older trunk version has at least two shards where the leader and the replica do not contain the same number of records. No recovery is attempted, it seems it thinks everything is alright. Also, one core of one of the unsynced shards waits forever loading /replication?command=detailwt=json, other cores load it in a few ms. Both cores of another unsynced shard does not show this problem. -Original message- From:John Nielsen j...@mcb.dk Sent: Fri 14-Dec-2012 11:50 To: solr-user@lucene.apache.org Subject: Re: Strange data-loss problem on one of our cores I did a manual commit, and we are still missing docs, so it doesn't look like the search race condition you mention. My boss wasn't happy when i mentioned that I wanted to try out unreleased code. Ill get him won over though and return with my findings. It will probably be some time next week. Thanks for your help. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Dec 13, 2012 at 4:10 PM, Mark Miller markrmil...@gmail.com wrote: Couple things to start: By default SolrCloud distributes updates a doc at a time. So if you have 1 shard, whatever node you index too, it will send updates to the other. Replication is only used for recovery, not distributing data. So for some reason, there is an IOException when it tries to forward. The other issue is not something that Ive seen reported. Can/did you try and do another hard commit to make sure you had the latest search open when checking the # of docs on each node? There was previously a race around commit that could cause some issues around expected visibility. If you are able to, you might try out a nightly build - 4.1 will be ready very soon and has numerous bug fixes for SolrCloud. - Mark On Dec 13, 2012, at 9:53 AM, John Nielsen j...@mcb.dk wrote: Hi all, We are seeing a strange problem on our 2-node solr4 cluster. This problem has resultet in data loss. We have two servers, varnish01 and varnish02
Strange data-loss problem on one of our cores
,+variant_of_item_guid+ascgroup.distributed.first=truefacet.limit=1000q.alt=*:*q.alt=*:*distrib=falsefacet.method=enumversion=2df=textfl=docidshard.url= varnish02.lynero.net:8000/solr/default1_Danish/|varnish01.lynero.net:8000/solr/default1_Danish/NOW=1355397828395group.field=groupby_variant_of_item_guidfacet.field=itemgroups_int_mvfq=site_guid:(11440)fq=item_type:(PRODUCT)fq=language_guid:(1)fq=item_group_56823_combination:(*)fq=item_group_45879_combination:(*)fq=is_searchable:(True)querytype=Technicalmm=100%25facet.missing=ongroup.ngroups=truefacet.mincount=1qf=%0a++text^0.5+name^1.2+searchable_text^0.8+typeahead_text^1.0+keywords^1.1+item_no^5.0%0a++ranking1_text^1.0+ranking2_text^2.0+ranking3_text^3.0%0a+++wt=javabingroup.facet=truedefType=edismaxrows=0facet.sort=lexstart=0group=truegroup.sort=name+ascisShard=true} status=0 QTime=8 Dec 13, 2012 12:23:48 PM org.apache.solr.common.cloud.ZkStateReader updateClusterState INFO: Updating cloud state from ZooKeeper... * Which is picked up on varnish01: *Dec 13, 2012 12:23:48 PM org.apache.solr.common.cloud.ZkStateReader$2 process INFO: A cluster state change has occurred - updating...* It looks like it replicated successfully, only it didnt. The default1_Norwegian core on varnish01 now has 55.071 docs and the same core on varnish02 has 35.088 docs. I checked the log files for both JVM's and no stop-the-world GC were taking place. There is also nothing in the zookeeper log of interest that I can see. -- Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: SOLR4 cluster - strange CPU spike on slave
Very interesting! I've seen references to NRTCachingDirectory, MMapDirectory, FSDirectory, RamDirectory and NIOFSDirectory, and thats just what I can remember. I have tried to search for more information about these, but I'm not having much luck. Is there a place where I can read up on these? Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Wed, Dec 5, 2012 at 1:11 AM, Mark Miller markrmil...@gmail.com wrote: On Dec 4, 2012, at 2:25 AM, John Nielsen j...@mcb.dk wrote: The post about mmapdirectory is really interesting. We switched to using that from NRTCachingDirectory and am monitoring performance as well. Initially performance doesn't look stellar, but i suspect that we lack memory in the server to really make it shine. NRTCachingDirectory delegates to another directory and simply caches small segments in RAM - usually it delegates MMapDirectory by default. So likely you won't notice any changes, because you have not likely really changed anything. NRTCachingDirectory simply helps in the NRT case and doesn't really hurt that I've seen in the std case. It's more like a help dir than a replacement. - Mark
Re: SOLR4 cluster - strange CPU spike on slave
I'm not sure I understand why this is important. Too much memory would just be unused. This is what the heap looks now: Heap Configuration: MinHeapFreeRatio = 40 MaxHeapFreeRatio = 70 MaxHeapSize = 17179869184 (16384.0MB) NewSize = 21757952 (20.75MB) MaxNewSize = 283508736 (270.375MB) OldSize = 65404928 (62.375MB) NewRatio = 7 SurvivorRatio= 8 PermSize = 21757952 (20.75MB) MaxPermSize = 176160768 (168.0MB) Heap Usage: New Generation (Eden + 1 Survivor Space): capacity = 255197184 (243.375MB) used = 108828496 (103.78694152832031MB) free = 146368688 (139.5880584716797MB) 42.644865548359654% used Eden Space: capacity = 226885632 (216.375MB) used = 83498424 (79.63030242919922MB) free = 143387208 (136.74469757080078MB) 36.80198841326365% used From Space: capacity = 28311552 (27.0MB) used = 25330072 (24.156639099121094MB) free = 2981480 (2.8433609008789062MB) 89.46903370044849% used To Space: capacity = 28311552 (27.0MB) used = 0 (0.0MB) free = 28311552 (27.0MB) 0.0% used concurrent mark-sweep generation: capacity = 16896360448 (16113.625MB) used = 12452710200 (11875.829887390137MB) free = 4443650248 (4237.795112609863MB) 73.70054775005708% used Perm Generation: capacity = 70578176 (67.30859375MB) used = 37652032 (35.90777587890625MB) free = 32926144 (31.40081787109375MB) 53.347981109627995% used Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Nov 29, 2012 at 4:08 AM, Otis Gospodnetic otis.gospodne...@gmail.com wrote: If this is caused by index segment merging you should be able to see that very clearly on the Index report in SPM, where you would see sudden graph movement at the time of spike and corresponding to CPU and disk activity. I think uncommenting that infostream in solrconfig would also show it. Otis -- SOLR Performance Monitoring - http://sematext.com/spm On Nov 28, 2012 9:20 PM, Erick Erickson erickerick...@gmail.com wrote: Am I reading this right? All you're doing on varnish1 is replicating to it? You're not searching or indexing? I'm sure I'm misreading this. The spike, which only lasts for a couple of minutes, sends the disks racing This _sounds_ suspiciously like segment merging, especially the disks racing bit. Or possibly replication. Neither of which make much sense. But is there any chance that somehow multiple commits are being issued? Of course if varnish1 is a slave, that shouldn't be happening either. And the whole bit about nothing going to the logs is just bizarre. I'm tempted to claim hardware gremlins, especially if you see nothing similar on varnish2. Or some other process is pegging the machine. All of which is a way of saying I have no idea Yours in bewilderment, Erick On Wed, Nov 28, 2012 at 6:15 AM, John Nielsen j...@mcb.dk wrote: I apologize for the late reply. The query load is more or less stable during the spikes. There are always fluctuations, but nothing on the order of magnitude that could explain this spike. In fact, the latest spike occured last night when there were almost noone using it. To test a hunch of mine, I tried to deactivate all caches by commenting out all cache entries in solrconfig.xml. It still spikes, so I dont think it has anything to do with cache warming or hits/misses or anything of the sort. One interesting thing GC though. This is our latest spike with cpu load (this server has 8 cores, so a load higher than 8 is potentially troublesome): 2012.Nov.27 19:58:182.27 2012.Nov.27 19:57:174.06 2012.Nov.27 19:56:188.95 2012.Nov.27 19:55:1719.97 2012.Nov.27 19:54:1732.27 2012.Nov.27 19:53:181.67 2012.Nov.27 19:52:171.6 2012.Nov.27 19:51:181.77 2012.Nov.27 19:50:171.89 This is what the GC was doing around that time: 2012-11-27T19:50:04.933+0100: [GC [PSYoungGen: 4777586K-277641K(4969216K)] 8887542K-4387693K(9405824K), 0.0856360 secs] [Times: user=0.54 sys=0.00, real=0.09 secs] 2012-11-27T19:50:30.785+0100: [GC [PSYoungGen: 4749769K-325171K(5068096K)] 8859821K-4435320K(9504704K), 0.0992160 secs] [Times: user=0.63 sys=0.00, real=0.10 secs] 2012-11-27T19:51:12.293+0100: [GC [PSYoungGen: 4911603K-306181K(5071168K)] 9021752K-4416617K(9507776K), 0.0957890 secs] [Times: user=0.62 sys=0.00, real=0.09 secs] 2012-11-27T19:51:52.817+0100: [GC [PSYoungGen: 4892613K-376175K(5075328K)] 9003049K-4486755K(9511936K), 0.1099830 secs] [Times: user=0.79 sys=0.01, real=0.11 secs] 2012-11-27T19:52:29.454+0100: [GC [PSYoungGen: 4972847K-271468K(4868160K)] 9083427K-4383520K(9304768K), 0.0699660 secs] [Times
Re: SOLR4 cluster - strange CPU spike on slave
Success! I tried adding -XX:+UseConcMarkSweepGC to java to make it GC earlier. We haven't seen any spikes since. I'm cautiously optimistic though and will be monitoring the servers for a week or so before declaring final victory. The post about mmapdirectory is really interesting. We switched to using that from NRTCachingDirectory and am monitoring performance as well. Initially performance doesn't look stellar, but i suspect that we lack memory in the server to really make it shine. Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Nov 30, 2012 at 3:13 PM, Erick Erickson erickerick...@gmail.comwrote: right, so here's what I'd check for. Your logs should show a replication pretty coincident with the spike and that should be in the log. Note: the replication should complete just before the spike. Or you can just turn replication off and fire it manually to try to force the situation at will, see: http://wiki.apache.org/solr/SolrReplication#HTTP_API. (but note that you'll have to wait until the index has changed on the master to see any action). So you should be able to create your spike at will. And this will be pretty normal. When replication happens, a new searcher is opened, caches are filled, autowarming is done, all kinds of stuff like that. During this period, the _old_ searcher is still open, which will both cause the CPU to be busier and require additional memory. Once the new searcher is warmed, new queries go to it, and when the old searcher has finished serving all the queries it shuts down and all the resources are freed. Which is why commits are expensive operations. All of which means that so far I don't think there's a problem, this is just normal Solr operation. If you're seeing responsiveness problems when serving queries you probably want to throw more hardware (particularly memory) at the problem. But when thinking about memory allocating to the JVM, _really_ read Uwe's post here: http://blog.thetaphi.de/2012/07/use-lucenes-mmapdirectory-on-64bit.html Best Erick On Thu, Nov 29, 2012 at 2:39 AM, John Nielsen j...@mcb.dk wrote: Yup you read it right. We originally intended to do all our indexing to varnish02, replicate to varnish01 and then search from varnish01 (through a fail-over ip which would switch the reader to varnish02 in case of trouble). When I saw the spikes, I tried to eliminate possibilities by starting searching from varnish02, leaving varnish01 with nothing to do but to receive replication data. This did not remove the spikes. As soon as this spike is fixed, I will start searching from varnish01 again. These sort of debug antics are only possible because, although we do have customers using this, we are still in our beta phase. Varnish01 never receives any manual commit orders. Varnish02 does from time to time. Oh, and I accidentally misinformed you before. (damn secondary language) We are actually seeing the spikes on both servers. I was just focusing on varnish01 because I use it to eliminate possibilities. It just occurred to me now; We tried switching off our feeder/index tool for 24 hours, and we didn't see any spikes during this period, so receiving replication data certainly has something to do with it. Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Nov 29, 2012 at 3:20 AM, Erick Erickson erickerick...@gmail.com wrote: Am I reading this right? All you're doing on varnish1 is replicating to it? You're not searching or indexing? I'm sure I'm misreading this. The spike, which only lasts for a couple of minutes, sends the disks racing This _sounds_ suspiciously like segment merging, especially the disks racing bit. Or possibly replication. Neither of which make much sense. But is there any chance that somehow multiple commits are being issued? Of course if varnish1 is a slave, that shouldn't be happening either. And the whole bit about nothing going to the logs is just bizarre. I'm tempted to claim hardware gremlins, especially if you see nothing similar on varnish2. Or some other process is pegging the machine. All of which is a way of saying I have no idea Yours in bewilderment, Erick On Wed, Nov 28, 2012 at 6:15 AM, John Nielsen j...@mcb.dk wrote: I apologize for the late reply. The query load is more or less stable during the spikes. There are always fluctuations, but nothing on the order of magnitude that could explain this spike. In fact, the latest spike occured last night when there were almost noone using it. To test a hunch of mine, I tried to deactivate all caches by commenting
Re: SOLR4 cluster - strange CPU spike on slave
this could be. Google has not given me any clues. Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Fri, Nov 23, 2012 at 1:56 PM, Otis Gospodnetic otis.gospodne...@gmail.com wrote: Strange indeed. What about query load/ayes during that time? What about GC? And does cache hit rate drop? Otis -- SOLR Performance Monitoring - http://sematext.com/spm On Nov 23, 2012 2:45 AM, John Nielsen j...@mcb.dk wrote: Hi all, We are seeing a strange CPU spike on one of our solr4 servers which we are unable to explain. The spike, which only lasts for a couple of minutes, sends the disks racing. This happens a few times a times a day. This is what the load looks like: 2012.Nov.14 13:37:172.77 2012.Nov.14 13:36:173.65 2012.Nov.14 13:35:183.92 2012.Nov.14 13:34:173.95 2012.Nov.14 13:33:186.56 2012.Nov.14 13:32:1710.79 2012.Nov.14 13:31:1724.38 2012.Nov.14 13:30:1763.35 2012.Nov.14 13:29:1724.68 2012.Nov.14 13:28:172.44 2012.Nov.14 13:27:183.51 2012.Nov.14 13:26:175.26 2012.Nov.14 13:25:185.71 2012.Nov.14 13:24:172.7 The problem is that out of a 3 minute spike, I get about 40 seconds of silence in the logs. This log usually adds like a thousand lines every second. Not being able to communicate with the server for this long, breaks our use case. We have two servers, varnish01 and varnish02. We used to feed data to varnish02, replicate it to varnish01 where the data is then read from. When we discovered this issue, we moved all traffic to varnish02 so that data is being replicated to varnish01, but other than that, gets zero traffic. The spike did not disappear. The spike we are seeing is on varnish01 only. Please note that our use case requires us to continuously feed large amounts of data from our main system in the order of up to 1.000 registers every minute. Our solrconfig.xml is attached. Has anyone seen this phenomenon before? Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk
Re: SOLR4 cluster - strange CPU spike on slave
Yup you read it right. We originally intended to do all our indexing to varnish02, replicate to varnish01 and then search from varnish01 (through a fail-over ip which would switch the reader to varnish02 in case of trouble). When I saw the spikes, I tried to eliminate possibilities by starting searching from varnish02, leaving varnish01 with nothing to do but to receive replication data. This did not remove the spikes. As soon as this spike is fixed, I will start searching from varnish01 again. These sort of debug antics are only possible because, although we do have customers using this, we are still in our beta phase. Varnish01 never receives any manual commit orders. Varnish02 does from time to time. Oh, and I accidentally misinformed you before. (damn secondary language) We are actually seeing the spikes on both servers. I was just focusing on varnish01 because I use it to eliminate possibilities. It just occurred to me now; We tried switching off our feeder/index tool for 24 hours, and we didn't see any spikes during this period, so receiving replication data certainly has something to do with it. Med venlig hilsen / Best regards *John Nielsen* Programmer *MCB A/S* Enghaven 15 DK-7500 Holstebro Kundeservice: +45 9610 2824 p...@mcb.dk www.mcb.dk On Thu, Nov 29, 2012 at 3:20 AM, Erick Erickson erickerick...@gmail.comwrote: Am I reading this right? All you're doing on varnish1 is replicating to it? You're not searching or indexing? I'm sure I'm misreading this. The spike, which only lasts for a couple of minutes, sends the disks racing This _sounds_ suspiciously like segment merging, especially the disks racing bit. Or possibly replication. Neither of which make much sense. But is there any chance that somehow multiple commits are being issued? Of course if varnish1 is a slave, that shouldn't be happening either. And the whole bit about nothing going to the logs is just bizarre. I'm tempted to claim hardware gremlins, especially if you see nothing similar on varnish2. Or some other process is pegging the machine. All of which is a way of saying I have no idea Yours in bewilderment, Erick On Wed, Nov 28, 2012 at 6:15 AM, John Nielsen j...@mcb.dk wrote: I apologize for the late reply. The query load is more or less stable during the spikes. There are always fluctuations, but nothing on the order of magnitude that could explain this spike. In fact, the latest spike occured last night when there were almost noone using it. To test a hunch of mine, I tried to deactivate all caches by commenting out all cache entries in solrconfig.xml. It still spikes, so I dont think it has anything to do with cache warming or hits/misses or anything of the sort. One interesting thing GC though. This is our latest spike with cpu load (this server has 8 cores, so a load higher than 8 is potentially troublesome): 2012.Nov.27 19:58:182.27 2012.Nov.27 19:57:174.06 2012.Nov.27 19:56:188.95 2012.Nov.27 19:55:1719.97 2012.Nov.27 19:54:1732.27 2012.Nov.27 19:53:181.67 2012.Nov.27 19:52:171.6 2012.Nov.27 19:51:181.77 2012.Nov.27 19:50:171.89 This is what the GC was doing around that time: 2012-11-27T19:50:04.933+0100: [GC [PSYoungGen: 4777586K-277641K(4969216K)] 8887542K-4387693K(9405824K), 0.0856360 secs] [Times: user=0.54 sys=0.00, real=0.09 secs] 2012-11-27T19:50:30.785+0100: [GC [PSYoungGen: 4749769K-325171K(5068096K)] 8859821K-4435320K(9504704K), 0.0992160 secs] [Times: user=0.63 sys=0.00, real=0.10 secs] 2012-11-27T19:51:12.293+0100: [GC [PSYoungGen: 4911603K-306181K(5071168K)] 9021752K-4416617K(9507776K), 0.0957890 secs] [Times: user=0.62 sys=0.00, real=0.09 secs] 2012-11-27T19:51:52.817+0100: [GC [PSYoungGen: 4892613K-376175K(5075328K)] 9003049K-4486755K(9511936K), 0.1099830 secs] [Times: user=0.79 sys=0.01, real=0.11 secs] 2012-11-27T19:52:29.454+0100: [GC [PSYoungGen: 4972847K-271468K(4868160K)] 9083427K-4383520K(9304768K), 0.0699660 secs] [Times: user=0.48 sys=0.01, real=0.07 secs] 2012-11-27T19:53:08.176+0100: [GC [PSYoungGen: 4868140K-336421K(5090944K)] 8980192K-4448572K(9527552K), 0.0824350 secs] [Times: user=0.56 sys=0.01, real=0.08 secs] 2012-11-27T19:54:53.534+0100: [GC [PSYoungGen: 4950373K-340513K(5092864K)] 9062524K-4468215K(9529472K), 0.1016770 secs] [Times: user=0.71 sys=0.00, real=0.10 secs] 2012-11-27T19:55:02.906+0100: [GC [PSYoungGen: 4954465K-480488K(4952000K)] 9082167K-4684537K(9388608K), 0.1813290 secs] [Times: user=1.23 sys=0.09, real=0.19 secs] 2012-11-27T19:55:09.114+0100: [GC [PSYoungGen: 4951976K-560434K(5031936K)] 9156025K-5075285K(9547072K), 0.3511090 secs] [Times: user=2.32 sys=0.12, real=0.35 secs] 2012-11-27T19:55:09.465+0100: [Full GC [PSYoungGen: 560434K-0K(5031936K)] [PSOldGen: 4514851K-2793342K(5047296K)] 5075285K-2793342K(10079232K) [PSPermGen: 35285K-35285K(44864K)], 5.2310820 secs
SOLR4 cluster - strange CPU spike on slave
Hi all, We are seeing a strange CPU spike on one of our solr4 servers which we are unable to explain. The spike, which only lasts for a couple of minutes, sends the disks racing. This happens a few times a times a day. This is what the load looks like: 2012.Nov.14 13:37:172.77 2012.Nov.14 13:36:173.65 2012.Nov.14 13:35:183.92 2012.Nov.14 13:34:173.95 2012.Nov.14 13:33:186.56 2012.Nov.14 13:32:1710.79 2012.Nov.14 13:31:1724.38 2012.Nov.14 13:30:1763.35 2012.Nov.14 13:29:1724.68 2012.Nov.14 13:28:172.44 2012.Nov.14 13:27:183.51 2012.Nov.14 13:26:175.26 2012.Nov.14 13:25:185.71 2012.Nov.14 13:24:172.7 The problem is that out of a 3 minute spike, I get about 40 seconds of silence in the logs. This log usually adds like a thousand lines every second. Not being able to communicate with the server for this long, breaks our use case. We have two servers, varnish01 and varnish02. We used to feed data to varnish02, replicate it to varnish01 where the data is then read from. When we discovered this issue, we moved all traffic to varnish02 so that data is being replicated to varnish01, but other than that, gets zero traffic. The spike did not disappear. The spike we are seeing is on varnish01 only. Please note that our use case requires us to continuously feed large amounts of data from our main system in the order of up to 1.000 registers every minute. Has anyone seen this phenomenon before? Best regards John Nielsen