nickva commented on code in PR #5870:
URL: https://github.com/apache/couchdb/pull/5870#discussion_r2747861382


##########
src/docs/src/maintenance/performance.rst:
##########
@@ -58,6 +58,14 @@ operations add the following to 
``(prefix)/etc/defaults/couchdb``
 
 .. _command line flag +A in erl(1): http://erlang.org/doc/man/erl.html
 
+Since Erlang OTP 21, the linked-in drivers have been migrated to dirty IO 
schedulers.
+These schedulers can be configured by ``ERL_FLAGS`` environment
+variable. For example, to give Erlang 80 dirty schedulers
+add the following to ``(prefix)/etc/defaults/couchdb``
+(or equivalent)::
+
+    export ERL_FLAGS="+Sdio 80"
+

Review Comment:
   > +Sdio 80"
   
   Note, the flag is `+SDio ...` (capital `D`) (double-check at 
https://www.erlang.org/doc/apps/erts/erl_cmd.html#emulator-flags)
   
   I wonder since the `+A` advice doesn't apply any longer maybe we should 
mention the correct way of updating the value and to use `vm.args` 
https://github.com/apache/couchdb/blob/main/rel/overlay/etc/vm.args#L71-L73. 
Maybe at the end, as minor note, we could add that this used to be an `+A` and 
that it doesn't apply any longer, and that `+A` setting should be removed.
   
   (suggestion)
   
   ---
   
   > ...However, there is one option that can increase the responsiveness of 
your CouchDB server when disk performance is a bottleneck, and that is setting 
the number dirty IO schedulers for the Erlang VM.
   
   > The setting is ``+SDio N`` where ``N ``is the number of dirty IO 
schedulers. CouchDB releases version 3.3.0 and higher set the value in the 
``vm.args`` file as ``+SDio 16``. Larger machines with more CPUs and faster 
disks could use a larger value like `+SDio 80`, for example.
   
   >These schedulers can also be configured by ``ERL_FLAGS`` environment 
variable. For example, to give Erlang 80 dirty schedulers add the following to 
``(prefix)/etc/defaults/couchdb`` (or equivalent)::
   
   >    export ERL_FLAGS="+SDio 80"
   ...
   
   ---
   
   That's just a suggestion, it can probably use better wording



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to