Pause serving sessions whose socket buffer is full (anti-DoS)
Serve requests in batches of 8
Don't store the session in the request
RPC has priority 0; every other session at least 1
Periodically consolidate small session groups into 1
Add new RPC command requests.
Adjust sessions RPC command to show pending requests.
Only keep first 15 chars in version string.
Set socket timeout
Try more forceful closing of socket
Fixes#67
Batch requests are now processed and encoded incrementally.
Their bandwidth usage is also incrementally added, and so overlarge
responses are rejected as soon as they become too large.
JSON text is also more memory efficient than the python data
structures they represent.
Show it in the logs with each session-specific message.
Show the sessions ID in the sessions list with an L suffix if logging.
Add RPC commands to toggle logging of, and disconnect, a session.
Closes#55
- incoming buffered network requests limited to 150,000 bytes
which I believe is ample for a genuine client
- if exceeded, the connection is dropped
- raise outgoing data limit for RPC connections to 5 MB - expect
sessions calls can be long and connection is implicitly trusted
- similarly raise incoming buffered data limit to 5 MB for RPC
connections
Move session logging code to protocol.py from electrum_rpc.py
Use it for periodic logging controlled by envvar LOG_SESSIONS
For each session, track sent transaction stats and show that
per-session instead of errors
The block processor needs to be able to close cleanly, and not
mid-block. In order to be able to yield whilst processing blocks
we cannot forcefully close its coroutine with a cancellation.
Neil Booth