Increase the timeout for RPC responses from Bitcoin Core #915
Conversation
Codecov Report
@@ Coverage Diff @@
## main #915 +/- ##
==========================================
- Coverage 90.51% 90.47% -0.04%
==========================================
Files 59 59
Lines 29771 29769 -2
==========================================
- Hits 26946 26934 -12
- Misses 2825 2835 +10
Continue to review full report at Codecov.
|
TheBlueMatt
force-pushed
the
2021-05-bump-rpc-timeout
branch
from
8da44a9 to
d0a738e
Compare
|
Fixed the comments and also added a new commit I'd missed to enable retrying if we actually hit a TCP socket timeout. |
| let read_res = reader.read_line(&mut line); | ||
| match read_res { | ||
| Ok(bytes_read) => break bytes_read, | ||
| Err(e) if e.kind() == std::io::ErrorKind::WouldBlock => { |
There was a problem hiding this comment.
Any other errors we would want to retry on? Like Interrupted docs https://doc.rust-lang.org/std/io/enum.ErrorKind.html#variant.Interrupted seems to indicate that it can probably be retried
There was a problem hiding this comment.
Hmm, I don't think you can get an Interrupted in a read on Linux and probably OSX. I'm not really sure what it would even mean in practice - usually it would imply another thread someone sent some signal to interrupt, but in a read not so much.
All that said, maybe it makes sense to just retry everything. Another potential error we could see is a Bitcoin Core RPC queue full error, which would present as just a generic HTTP error.
TheBlueMatt
force-pushed
the
2021-05-bump-rpc-timeout
branch
from
d0a738e to
3d3379d
Compare
|
Redid the first commit to always retry requests instead of checking the error source. |
Early sample testing showed multiple users hitting EWOULDBLOCK/EAGAIN waiting for an initial response from Bitcoin Core while it was doing some long operation (eg UTXO cache flushing). Instead of only waiting 5 seconds for each attempt, we now wait a full two minutes, but only for the first header response, not each byte.
TheBlueMatt
force-pushed
the
2021-05-bump-rpc-timeout
branch
from
3d3379d to
4ade6bc
Compare
|
Squashed fixup commits with no change: |
Early sample testing showed multiple users hitting
EWOULDBLOCK/EAGAIN waiting for an initial response from Bitcoin
Core while it was doing some long operation (eg UTXO cache
flushing). Instead of only waiting 5 seconds for each attempt, we
now wait a full two minutes, but only for the first header
response, not each byte.