On 3/7/2018 12:10 PM, Bsr wrote:
I guess i should increase autowarmCount count. whats should be the ideal no.
Also is there any way by which i can know that autowarm is completed?
There are no generic answers. You want autowarmCount to be large enough
to be effective, but small enough that warming doesn't take a really
long time. There's no way I can tell you how long warming will take
with a certain number in autowarmCount. That will depend on the nature
of your queries, what's in your index, and what the hardware is.
When the new searcher opens, that's when you will know that all warming
is complete. The cache stats will show how long it took for for
aspecific cache to warm up when that instance of the cache was created.
In another message you askedthis:
Can you eloborate more on newSearcher and cache i.e what should we set.
The newSearcher config defines queries to execute on *every* new
searcher. Which is different than firstSearcher, in that firstSearcher
defines queries that will be executed exactly once, when the core first
starts up. If you want to use newSearcher, it probably needs the same
queries you currently have in firstSearcher. If this is a warming
issue, adding newSearcher will probably help.
The caches themselves have autowarming, which reads the top N queries
from the cache from the old searcher and re-executes those queries on
the new index to populate the cache in the new searcher. This tends to
produce better results than newSearcher, because the queries might be
different, and will reflect what's actually IN the cache.
You said this as well: Also there is no resource crunch with the solr
resources.
At the risk of being offensive: How do you know? I find that many
people do not actually know how to detect resource issues with Solr,
particularly with memory. They look at their systems and conclude that
everything's fine, even though there is nowhere near enough memory
installed in the system for good performance.
Thanks,
Shawn
