Hi! For uploading the scripts from inside the image, I didn't implement a locking mechanism or assigned an instance 🤔 Since if configset was already uploaded on the first ever run, on the subsequent runs it would just log an error (configset already exists) and continue with the rest of the process. I don't know if two Solr pods can start exactly at the same time on the first ever run and conflict though. My guess would be that Zookeeper itself would handle that situation.
-ufuk yilmaz ________________________________ From: Дмитрий Фитисов <[email protected]> Sent: Tuesday, January 30, 2024 2:02 PM To: [email protected] <[email protected]> Subject: Re: Configuring/reconfiguring k8s SolrCloud cluster GitOps way Thank you! I’ve looked at Solr Operator and seems like it’s a good way to go. Regarding the upload from the script inside an image, did you implement some kind of a locking mechanism or “assigned” one of the instances to do the job? Best regards, Dmitrii Fitisov On Tue, 30 Jan 2024 at 11:54, ufuk yılmaz <[email protected]> wrote: > Hi! There is an official Solr project to manage Solr on kubernetes called > “Solr Operator”. You can also check it if you wish, instead of rolling your > own implementation. > > Alternatively I used to build a custom docker image based on Solr image > and put configsets in there, and upload them on Solr startup via scripts in > ‘docker-entrypoint-initdb.d’ folder. That way I could put both image and > configsets in version control. > > —ufuk yilmaz > — > > > On 30 Jan 2024, at 13:24, Дмитрий Фитисов <[email protected]> wrote: > > > > Hello there! I've been researching moving from Standalone Solr to k8s > > SolrCloud setup and I'm wondering how you guys manage configsets and > > collections when operating in k8s? It seems to me that the easiest > > solution is to introduce some kind of an admin process (In a form of a > > bash script) that will upload configsets to zookeeper via the call to > > a zookeeper service and reload/create collections via the call to a > > solr service. The process itself should be spawned after all the Solr > > pods (and Zookeeper?) are live and ready. > > > > Can you share your way of dealing with this?Maybe I'm missing something? > > > > Best regards, > > Dmitrii Fitisov > >
