There may be other ways, easiest way is to write a script that gets the cluster status, and for each collection per replica you will have these details:
"collections":{ “collection1":{ "pullReplicas":"0", "replicationFactor":"1", "shards":{ "shard1":{ "range":"80000000-8ccbffff", "state":"active", "replicas":{"core_node33":{ "core”:"collection1_shard1_replica_n30", "base_url":"http://host:port/solr", "node_name”:”host:port", "state":"active", "type":"NRT", "force_set_state":"false", "leader":"true"}}}, For each replica of each shard make a localized call for numRecords: base_url/core/sleect?q=*:*&shard=shardX&distrib=false&rows=0 If you have replicas that disagree with each other with the number of records per shard then u have an issue with replicas not being in sync for a collection. This is what I meant when I said “replicas out of sync”. Your situation was actually very simple :) one of you collections has less data. You seem to have a sync requirement between collections which is interesting, but thats beyond solr. Your inter collection sync script needs some debugging most likely :) > On Aug 12, 2020, at 4:29 PM, Jae Joo <jaejo...@gmail.com> wrote: > > Good question. How can I validate if the replicas are all synched? > > > On Wed, Aug 12, 2020 at 7:28 PM Jae Joo <jaejo...@gmail.com> wrote: > >> numFound is same but different score. >> <result name="response" numFound="755970" start="0" maxScore="4.70519"> >> <result name="response" numFound="755970" start="0" maxScore="4.70519"> >> <result name="response" numFound="755970" start="0" maxScore="4.70519"> >> <result name="response" numFound="755970" start="0" maxScore="4.7738605"> >> <result name="response" numFound="755970" start="0" maxScore="4.659804"> >> <result name="response" numFound="755970" start="0" maxScore="4.659804"> >> <result name="response" numFound="755970" start="0" maxScore="4.659804"> >> >> On Wed, Aug 12, 2020 at 6:01 PM Aroop Ganguly >> <aroopgang...@icloud.com.invalid> wrote: >> >>> Try a simple test of querying each collection 5 times in a row, if the >>> numFound are different for a single collection within tase 5 calls then u >>> have it. >>> Please try it, what you may think is sync’d may actually not be. How do >>> you validate correct sync ? >>> >>>> On Aug 12, 2020, at 10:55 AM, Jae Joo <jaejo...@gmail.com> wrote: >>>> >>>> The replications are all synched and there are no updates while I was >>>> testing. >>>> >>>> >>>> On Wed, Aug 12, 2020 at 1:49 PM Aroop Ganguly >>>> <aroopgang...@icloud.com.invalid> wrote: >>>> >>>>> Most likely you have 1 or more collections behind the alias that have >>>>> replicas out of sync :) >>>>> >>>>> Try querying each collection to find the one out of sync. >>>>> >>>>>> On Aug 12, 2020, at 10:47 AM, Jae Joo <jaejo...@gmail.com> wrote: >>>>>> >>>>>> I have 10 collections in single alias and having different result sets >>>>> for >>>>>> every time with the same query. >>>>>> >>>>>> Is it as designed or do I miss something? >>>>>> >>>>>> The configuration and schema for all 10 collections are identical. >>>>>> Thanks, >>>>>> >>>>>> Jae >>>>> >>>>> >>> >>>