This benchmark is designed to stress test the rpc methods of a lotus node so that we can simulate real world usage and measure the performance of rpc methods on the node.
This benchmark has the following features:
* Can query each method both sequentially and concurrently
* Supports rate limiting
* Can query multiple different endpoints at once (supporting different concurrency level and rate limiting for each method)
* Gives a nice reporting summary of the stress testing of each method (including latency distribution, histogram and more)
* Easy to use
To use this benchmark you must specify the rpc methods you want to test using the --method options, the format of it is:
--method=NAME[:CONCURRENCY][:QPS][:PARAMS] where only METHOD is required.
Here are some real examples:
lotus-bench rpc --method='eth_chainId' // run eth_chainId with default concurrency and qps
lotus-bench rpc --method='eth_chainId:3' // override concurrency to 3
lotus-bench rpc --method='eth_chainId::100' // override to 100 qps while using default concurrency
lotus-bench rpc --method='eth_chainId:3:100' // run using 3 workers but limit to 100 qps
lotus-bench rpc --method='eth_getTransactionCount:::["0xd4c70007F3F502f212c7e6794b94C06F36173B36", "latest"]' // run using optional params while using default concurrency and qps
lotus-bench rpc --method='eth_chainId' --method='eth_getTransactionCount:10:0:["0xd4c70007F3F502f212c7e6794b94C06F36173B36", "latest"]' // run multiple methods at once`,
Fixes: https://github.com/filecoin-project/lotus/issues/10752
This PR introduces as sharded mutex within the ChainIndex#GetTipsetByHeight.
It also replaces a go map with xsync.Map which doesn't require locking.
The lock is taken when it appears that ChainIndex filling work should be
started. After claiming the lock, the status of the cache is rechecked,
if the entry is still missing, the fillCache is started.
Thanks to @snissn and @arajasek for debugging and taking initial stabs at this.
Supersedes #10866 and 10885
Signed-off-by: Jakub Sztandera <kubuxu@protocol.ai>
