I'm ambivalent on the standalone vs cloud default.  I think this is in part
being driven by some features requiring SolrCloud when they needn't be tied
to it.  Just one example is Distributed Tracing support, which I recently
corrected (although it wasn't my primary goal).  Also the Package Manager.
Granted for that and other features, ZooKeeper is playing a key role, but
sometimes our features could be made to work reasonably without it.

RE embedded ZooKeeper:  On the user list, I recently posed a question to
our users about Solr security[1].  In that message, I suggested a new
env=dev|prod setting that would allow Solr to pick defaults that make sense
for a developer mode vs production mode.  Such a setting wouldn't strictly
be about security settings; it's really anything where you'd want the
setting to vary between dev or prod.  Embedded Zookeeper might be fine in
dev mode.

[1]
https://lists.apache.org/thread.html/re7499d64222377f169b81f4b2d46fcf9725609f585efa2566097f749%40%3Cusers.solr.apache.org%3E

~ David Smiley
Apache Lucene/Solr Search Developer
http://www.linkedin.com/in/davidwsmiley


On Mon, May 24, 2021 at 4:17 PM Anshum Gupta <[email protected]> wrote:

> I'll wait for Houston to clarify what he meant with his suggestion. My
> interpretation of that was that we default to cloud mode i.e. users get
> more OTB features but also we're almost nudging them in that direction.
>
> We could default to "-c" without having to explicitly type it out, and
> have a warning at startup but of course, we're assuming that they'd read
> those warnings and startup messages.
>
> 1. Default to -c, and
> 2. Make one of the following mandatory
>     i. -z for external zk
>    ii. an option to start solr in dev mode (embedded zk), so we also
> highlight that it is a non-production ready mode that's meant to make the
> getting started experience better.
>   iii. -standalone (or something else) that explicitly says they want to
> run w/o ZK, in which case print out a warning that highlights the lack of
> the distributed features in that mode.
>
>
>
>
> On Mon, May 24, 2021 at 12:52 PM Chris Hostetter <[email protected]>
> wrote:
>
>>
>> : Clearly this won't be done in 8.x. I think it would be ok to make the
>> : switch in Solr 9, but maybe it's something we aim for Solr 10 instead.
>> In
>> : the meantime we should probably include a backwards-compatible
>> standalone
>> : flag (-l/-standalone) that will start to be used once the default mode
>> is
>> : switched.
>>
>> It's not clear to me what exactly you are suggesting the new "default"
>> option should do w/regard to ZooKeeper?
>>
>> Jan's response seems to assume that you are suggesting that the default
>> behavior in Solr 9.x shold be equivilent to 8.x's "-c" (w/o a "-z"
>> option)
>> so that solr starts an embedded zookeeper ... this seems like a really
>> bad
>> idea to me, because it would "normalize" and encourage using the (non
>> fault tollerant) embedded zookeeper even more then it already is in 8.x
>> ... not to mention that anyone trying to *ACTUALLY* run a (mult-node)
>> "solr cluster" will still have to use "-z PORT_FROM_FIRST_SOLR_NODE" ...
>> at least right now with the mandatory "-c" option anyone looking for how
>> to start up a "cluster" can see the docs for "-c" pointing out hte
>> importance of "-z"
>>
>> On the otherhand: A differnet way of interpreting your suggestion is that
>> if we're changing the command line args, and eliminating "-c" (because ow
>> it's the default) we can start making "-z" required unless you specifying
>> "-standalone" ... or some new "-run-single-point-of-failure-embedded-zk"
>> so that solr defaults to cloud mode, but also deaults to failing fast if
>> you don't have a zk cluster running (or know that you are doing something
>> unsupported in production)
>>
>> Jan's interpretation of your suggestion scares me, but i could get on
>> board the second interpretation of pushing forward cloud mode and trying
>> to make people more aware of the importance of a zk cluster.
>>
>> (W/o more concrete specifics, I'm sure there are other interpretations of
>> this suggestion as well, these are just the 2 that immediately popped to
>> mind)
>>
>>
>>
>> -Hoss
>> http://www.lucidworks.com/
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: [email protected]
>> For additional commands, e-mail: [email protected]
>>
>>
>
> --
> Anshum Gupta
>

Reply via email to