I don't think so. I'd rather bark on distribute search reduce phase, if you use one. I don't know.
On Wed, Jun 1, 2022 at 6:58 AM Poorna Murali <poornamur...@gmail.com> wrote: > Thanks Mikhail! Regarding the id field, is it possible that while doing > faceting or sorting, the id fields will be loaded in field cache by > default? > On 2022/05/31 21:27:57 Mikhail Khludnev wrote: > > Hi, Poorna. > > Pls check inline below. > > > > On Tue, May 31, 2022 at 4:21 PM Poorna Murali <po...@gmail.com> > > wrote: > > > > > Hi, > > > > > > We are planning to introduce docValues and reduce the usage of > fieldcache > > > in our application. > > > > > > I have a few doubts on these. Please help me to clarify them. > > > > > > 1) Will there be a performance impact if we use docValues as it does > I/O > > > read? Especially at the time of index generation? > > > > > Sure, it needs to write more files during indexing. Then, query > > performance depends on particular IO. It definitely make sense to limit > jvm > > heap leaving some RAM free for memory mapping index files. > > > > > > > > > > 2) Are there any list of recommended fields/ category of fields that > are > > > suitable for docValues? > > > > > Short enums and numbers. Texts (in terms of analysis) are not supported > > anyway. > > > > > > > > 3) There is a field named "id" which is used as unique key in solr. > > > Although it's not used for faceting or sorting, it still comes up in > > > fieldcache and occupies lot of memory. Please help me understand how > this > > > id field came in field cache. > > > > > Some internal routines can trigger its loading. I'm afraid only debugger > > can give a definite answer. Maybe declaring it as uninvertible=false > can > > find a code path, but I'm not sure. > > > > > > > > 4) I understand that by looking to field cache in solr admin page, we > will > > > get the current entries in it. But, is it possible to get the whole > list > > > of attributes that can go into field cache? > > > > > I don't think it's possible, since field cache is lazy/runtime. ie one > can > > trigger new field caching by requesting a facet over it. > > > > > > > > Thanks in advance, > > > Poorna > > > > > Welcome! > > > > > > -- > > Sincerely yours > > Mikhail Khludnev > > > -- Sincerely yours Mikhail Khludnev
