Api middlewares refactor #2092
Api middlewares refactor #2092
Conversation
- added reset call failure watchdog in middleware
- implemented closing on nodeFacade - updated node's json files
…etwork/elrond-go into api-middlewares-refactor
| } | ||
|
|
||
| select { | ||
| case m.queue <- struct{}{}: |
There was a problem hiding this comment.
Interesting way of achieving requests throttling. Could also have been achieved using an atomic counter perhaps and dropping the request immediately. If I understand correctly, this logic also drops the requests on default, when the buffered channel is full. (I am thinking out loud).
| "address": "erd12ytze3088nhdp45rkm4k7ycvqeprlc7mvm6duu52rtzggp0allasw53ndn", | ||
| "supply": "1667291666666666666666666674", | ||
| "balance": "1664791666666666666666666674", | ||
| "address": "erd1ulhw20j7jvgfgak5p05kv667k5k9f320sgef5ayxkt9784ql0zssrzyhjp", |
There was a problem hiding this comment.
Please add in the PR description this change as well.
| @@ -11,40 +11,40 @@ | |||
| "adaptivity": false, | |||
| "initialNodes": [ | |||
| { | |||
| "pubkey": "153dae6cb3963260f309959bf285537b77ae16d82e9933147be7827f7394de8dc97d9d9af41e970bc72aecb44b77e819621081658c37f7000d21e2d0e8963df83233407bde9f46369ba4fcd03b57f40b80b06c191a428cfb5c447ec510e79307", | |||
| "address": "erd1qqqqqqqqqqqqqpgq73ejktqxvhgem304gxgsfa5gtcguh8p3wt8sk7nsql" | |||
| "pubkey": "cbc8c9a6a8d9c874e89eb9366139368ae728bd3eda43f173756537877ba6bca87e01a97b815c9f691df73faa16f66b15603056540aa7252d73fecf05d24cd36b44332a88386788fbdb59d04502e8ecb0132d8ebd3d875be4c83e8b87c55eb901", | |||
There was a problem hiding this comment.
Please add in the PR description this changes as well.
| log.Trace("calling reset on WS source limiter") | ||
| reset.Reset() | ||
| select { | ||
| case <-time.After(betweenResetDuration): |
There was a problem hiding this comment.
So this is a "hard" reset, which means from time to time we forgive everyone, which means that an eventual attacker, if starting right before the reset and continuing immediately after a reset, could send maxNumRequestsPerAddress * 2 requests in a short period of time (just thinking out loud, shouldn't be an issue, sliding mechanism not really required).
| @@ -343,6 +346,14 @@ func (nf *nodeFacade) GetThrottlerForEndpoint(endpoint string) (core.Throttler, | |||
| return throttlerForEndpoint, isThrottlerOk | |||
| } | |||
|
|
|||
| // Close will cleanup started go routines | |||
| // TODO use this close method | |||
There was a problem hiding this comment.
Perhaps add a Jira ticket for this, so that we don't forget?
c.Next()on the gin context multiple times