This repository has been archived by the owner on Nov 6, 2020. It is now read-only.
up to 30 seconds delay to rpc response #8746
Labels
F3-annoyance 💩
The client behaves within expectations, however this “expected behaviour” itself is at issue.
M6-rpcapi 📣
RPC API.
P5-sometimesoon 🌲
Issue is worth doing soon.
Milestone
Comments
Tbaut
added
P5-sometimesoon 🌲
|
we have exactly same issue, 4 nodes behind load balancer on Digital Ocean with significant traffic being handled nicely EXCEPT |
|
Facing the same issue, |
debris
added a commit
that referenced
this issue
Sep 12, 2018
5chdn
added a commit
that referenced
this issue
Sep 17, 2018
* parity-version: mark 2.0.5 track stable * deps: bump fs-swap to 0.2.4 * Remove initial token for WS. (#9545) * version: mark release critical * Increase Gas-floor-target and Gas Cap (#9564) + Gas-floor-target increased to 8M by default + Gas-cap increased to 10M by default * Improve P2P discovery (#9526) * Add `target` to Rust traces * network-devp2p: Don't remove discovery peer in main sync * network-p2p: Refresh discovery more often * Update Peer discovery protocol * Run discovery more often when not enough nodes connected * Start the first discovery early * Update fast discovery rate * Fix tests * Fix `ping` tests * Fixing remote Node address ; adding PingPong round * Fix tests: update new +1 PingPong round * Increase slow Discovery rate Check in flight FindNode before pings * Add `deprecated` to deprecated_echo_hash * Refactor `discovery_round` branching * net_version caches network_id to avoid redundant aquire of sync read lock (#9544) * net_version caches network_id to avoid redundant aquire of sync read lock, #8746 * use lower_hex display formatting for net_peerCount rpc method
5chdn
added a commit
that referenced
this issue
Sep 17, 2018
* parity-version: mark 2.1.0 track beta * ci: update branch version references * docker: release master to latest * Fix checkpointing when creating contract failed (#9514) * ci: fix json docs generation (#9515) * fix typo in version string (#9516) * Update patricia trie to 0.2.2 crates. Default dependencies on minor version only. * Putting back ethereum tests to the right commit * Enable all Constantinople hard fork changes in constantinople_test.json (#9505) * Enable all Constantinople hard fork changes in constantinople_test.json * Address grumbles * Remove EIP-210 activation * 8m -> 5m * Temporarily add back eip210 transition so we can get test passed * Add eip210_test and remove eip210 transition from const_test * In create memory calculation is the same for create2 because the additional parameter was popped before. (#9522) * deps: bump fs-swap and kvdb-rocksdb * Multithreaded snapshot creation (#9239) * Add Progress to Snapshot Secondary chunks creation * Use half of CPUs to multithread snapshot creation * Use env var to define number of threads * info to debug logs * Add Snapshot threads as CLI option * Randomize chunks per thread * Remove randomness, add debugging * Add warning * Add tracing * Use parity-common fix seek branch * Fix log * Fix tests * Fix tests * PR Grumbles * PR Grumble II * Update Cargo.lock * PR Grumbles * Default snapshot threads to half number of CPUs * Fix default snapshot threads // min 1 * correct before_script for nightly build versions (#9543) - fix gitlab array of strings syntax error - get proper commit id - avoid colon in stings * Remove initial token for WS. (#9545) * version: mark release critical * ci: fix rpc docs generation 2 (#9550) * Improve P2P discovery (#9526) * Add `target` to Rust traces * network-devp2p: Don't remove discovery peer in main sync * network-p2p: Refresh discovery more often * Update Peer discovery protocol * Run discovery more often when not enough nodes connected * Start the first discovery early * Update fast discovery rate * Fix tests * Fix `ping` tests * Fixing remote Node address ; adding PingPong round * Fix tests: update new +1 PingPong round * Increase slow Discovery rate Check in flight FindNode before pings * Add `deprecated` to deprecated_echo_hash * Refactor `discovery_round` branching * net_version caches network_id to avoid redundant aquire of sync read lock (#9544) * net_version caches network_id to avoid redundant aquire of sync read lock, #8746 * use lower_hex display formatting for net_peerCount rpc method * Increase Gas-floor-target and Gas Cap (#9564) + Gas-floor-target increased to 8M by default + Gas-cap increased to 10M by default * Revert to old parity-tokio-ipc. * Downgrade named pipes.
|
|
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
I'm running:
I'm running parity with nginx as a reverse tls-proxy on Digital Ocean instance with ssd disks 8GB ram, 4 vcpu.
It's fully synced, but I ran rsync to copy db from local parity.
Running 1800 requests using Chrome. For some reason I got up to 30 seconds delays to some "easy" requests with no shortage on io/ram/cpu as docker stats/ iostat / top tells me.
parity is using ~550 MB running for several days with no limit set in docker.
This request always tops responce time {jsonrpc: "2.0", id: 249, method: "net_version", params: []}.
But production parity v1.9.7 running in docker on VERY slow disks in azure and getting desynced sometimes have much better average and worst responce times with iowait for 20-30%, BUT using 10GB+ memory.
I can test 1.9.7 on DO, but I suppose 10GB is some db cache that makes parity reply much faster, what I am I doing wrong here?
The text was updated successfully, but these errors were encountered: