interleave read and write on pipelined_get (#776, #941) #942
Conversation
marvinthepa
force-pushed
the
interleaved_fetch
branch
3 times, most recently
from
350d7b4 to
03ad88e
Compare
|
There's a bit of "devil in the details", but directionally this looks correct. Thanks for drafting @marvinthepa . With a little back and forth we can almost certainly get this in. |
Happy to give back and spend some company time on a library we have been using for a few years now.. |
marvinthepa
force-pushed
the
interleaved_fetch
branch
from
03ad88e to
dac656d
Compare
marvinthepa
force-pushed
the
interleaved_fetch
branch
from
dac656d to
1030225
Compare
marvinthepa
force-pushed
the
interleaved_fetch
branch
3 times, most recently
from
427f027 to
6493f3d
Compare
marvinthepa
changed the title
|
@petergoldstein sorry for not following up on this PR for so long, but I found some time and hope I can manage to get this ready to integrate in the next weeks. I rebased on main, fixing the conflicts with especially #963 - so maybe if @casperisfine could check this PR as well it would be highly appreciated. I am still not 100% confident if the error handling is completely correct, but I was not able to find a flaw (at least not one that did not exist before anyway) after some time searching for one. |
fixes petergoldstein#776. fixes petergoldstein#941. When reading a large number of keys, memcached starts sending the response when dalli is not yet finished sending the request. As we did not start reading the response until we were finished writing the request, this could lead to the following problem: * the receive buffer (rcvbuf) would fill up * due to TCP backpressure, memcached would stop sending (okay as we are not reading anyway), but also stop reading * the send buffer (sndbuf) would also fill up * as we were using a blocking write without timeout, we would block forever (at least with ruby < 3.2, which introduces IO::Timeout, see petergoldstein#967) This is addressed by using IO::select on the sockets for both read and write, and thus start reading as soon as data is available.
marvinthepa
force-pushed
the
interleaved_fetch
branch
from
6493f3d to
f20883f
Compare
|
Back to draft, as I think I found a problem relating to #963, where the connection is unnecessarily closed during a pipelined get. sigh. |
fixes #776.
fixes #941.
When reading a large number of keys, memcached starts sending the
response when dalli is not yet finished sending the request.
As we did not start reading the response until we were finished writing
the request, this could lead to the following problem:
not reading anyway), but also stop reading
forever (at least with ruby < 3.2, which introduces IO::Timeout, see
IO::TimeoutError exceptions are not handled, breaking automatic retry #967)
This is addressed by using IO::select on the sockets for both read and
write, and thus start reading as soon as data is available.