improve health check behavior
This commit is contained in:
@@ -8,12 +8,14 @@ import {
|
||||
View,
|
||||
} from '@opentelemetry/sdk-metrics-base';
|
||||
import {
|
||||
ORDERBOOK_UPDATE_INTERVAL,
|
||||
commitHash,
|
||||
driftEnv,
|
||||
endpoint,
|
||||
getSlotHealthCheckInfo,
|
||||
wsEndpoint,
|
||||
} from '..';
|
||||
import { E_ALREADY_LOCKED, tryAcquire } from 'async-mutex';
|
||||
|
||||
/**
|
||||
* Creates {count} buckets of size {increment} starting from {start}. Each bucket stores the count of values within its "size".
|
||||
@@ -116,53 +118,73 @@ const responseStatusCounter = meter.createCounter(
|
||||
}
|
||||
);
|
||||
|
||||
const healthCheckInterval = 2000;
|
||||
const healthCheckInterval = 2 * (ORDERBOOK_UPDATE_INTERVAL ?? 1000); // ORDERBOOK_UPDATE_INTERVAL is NaN here for some reason ... hardcode to 1000.
|
||||
let lastHealthCheckSlot = -1;
|
||||
let lastHealthCheckSlotUpdated = Date.now();
|
||||
let lastHealthCheckState = true; // true = healthy, false = unhealthy
|
||||
let lastHealthCheckPerformed = Date.now() - healthCheckInterval;
|
||||
/**
|
||||
* Middleware that checks if we are in general healthy by checking that the bulk account loader slot
|
||||
* has changed recently.
|
||||
*
|
||||
* We may be hit by multiple sources performing health checks on us, so this middleware will latch
|
||||
* to its health state and only update every `healthCheckInterval`.
|
||||
*/
|
||||
const handleHealthCheck = async (req, res, next) => {
|
||||
const { lastSlotReceived, lastSlotReceivedMutex } = getSlotHealthCheckInfo();
|
||||
|
||||
try {
|
||||
if (req.url === '/health' || req.url === '/') {
|
||||
// check if a slot was received recently
|
||||
let healthySlotSubscriber = false;
|
||||
let slotChanged = false;
|
||||
let slotChangedRecently = false;
|
||||
await lastSlotReceivedMutex.runExclusive(async () => {
|
||||
slotChanged = lastSlotReceived > lastHealthCheckSlot;
|
||||
slotChangedRecently =
|
||||
Date.now() - lastHealthCheckSlotUpdated < healthCheckInterval;
|
||||
healthySlotSubscriber = slotChanged || slotChangedRecently;
|
||||
logger.debug(
|
||||
`Slotsubscriber health check: lastSlotReceived: ${lastSlotReceived}, lastHealthCheckSlot: ${lastHealthCheckSlot}, slotChanged: ${slotChanged}, slotChangedRecently: ${slotChangedRecently}`
|
||||
);
|
||||
if (slotChanged) {
|
||||
lastHealthCheckSlot = lastSlotReceived;
|
||||
lastHealthCheckSlotUpdated = Date.now();
|
||||
}
|
||||
});
|
||||
if (!healthySlotSubscriber) {
|
||||
healthStatus = HEALTH_STATUS.UnhealthySlotSubscriber;
|
||||
logger.error(`SlotSubscriber is not healthy`);
|
||||
logger.error(
|
||||
`Slotsubscriber health check: lastSlotReceived: ${lastSlotReceived}, lastHealthCheckSlot: ${lastHealthCheckSlot}, slotChanged: ${slotChanged}, slotChangedRecently: ${slotChangedRecently}`
|
||||
);
|
||||
|
||||
res.writeHead(500);
|
||||
res.end(`SlotSubscriber is not healthy`);
|
||||
return;
|
||||
}
|
||||
|
||||
// liveness check passed
|
||||
healthStatus = HEALTH_STATUS.Ok;
|
||||
if (Date.now() < lastHealthCheckPerformed + healthCheckInterval) {
|
||||
if (lastHealthCheckState) {
|
||||
res.writeHead(200);
|
||||
res.end('OK');
|
||||
} else {
|
||||
res.writeHead(404);
|
||||
res.end('Not found');
|
||||
res.writeHead(500);
|
||||
res.end(`NOK`);
|
||||
}
|
||||
return;
|
||||
}
|
||||
|
||||
const { lastSlotReceived, lastSlotReceivedMutex } = getSlotHealthCheckInfo();
|
||||
|
||||
try {
|
||||
await tryAcquire(lastSlotReceivedMutex).runExclusive(async () => {
|
||||
// healthy if slot has advanced since the last check
|
||||
lastHealthCheckState = lastSlotReceived > lastHealthCheckSlot;
|
||||
if (!lastHealthCheckState) {
|
||||
logger.error(
|
||||
`Unhealthy: lastSlot: ${lastSlotReceived}, lastHealthCheckSlot: ${lastHealthCheckSlot}`
|
||||
);
|
||||
}
|
||||
|
||||
lastHealthCheckSlot = lastSlotReceived;
|
||||
lastHealthCheckPerformed = Date.now();
|
||||
});
|
||||
|
||||
if (!lastHealthCheckState) {
|
||||
healthStatus = HEALTH_STATUS.UnhealthySlotSubscriber;
|
||||
|
||||
res.writeHead(500);
|
||||
res.end(`NOK`);
|
||||
return;
|
||||
}
|
||||
|
||||
healthStatus = HEALTH_STATUS.Ok;
|
||||
res.writeHead(200);
|
||||
res.end('OK');
|
||||
} catch (e) {
|
||||
next(e);
|
||||
if (e === E_ALREADY_LOCKED) {
|
||||
// someone else is updating the slot, just return the latched state
|
||||
if (lastHealthCheckState) {
|
||||
logger.error(`DlobServer is healthy (busy mutex)`);
|
||||
res.writeHead(200);
|
||||
res.end('OK');
|
||||
} else {
|
||||
logger.error(`DlobServer is not healthy (busy mutex)`);
|
||||
res.writeHead(500);
|
||||
res.end(`NOK`);
|
||||
}
|
||||
} else {
|
||||
logger.error(`Error in health check: ${e.message}`);
|
||||
console.error(e);
|
||||
next(e);
|
||||
}
|
||||
}
|
||||
};
|
||||
|
||||
|
||||
@@ -52,7 +52,7 @@ const sdkConfig = initialize({ env: process.env.ENV });
|
||||
|
||||
const stateCommitment: Commitment = 'processed';
|
||||
const serverPort = process.env.PORT || 6969;
|
||||
const ORDERBOOK_UPDATE_INTERVAL = 1000;
|
||||
export const ORDERBOOK_UPDATE_INTERVAL = 1000;
|
||||
|
||||
const rateLimitCallsPerSecond = process.env.RATE_LIMIT_CALLS_PER_SECOND
|
||||
? parseInt(process.env.RATE_LIMIT_CALLS_PER_SECOND)
|
||||
@@ -205,9 +205,7 @@ const main = async () => {
|
||||
});
|
||||
|
||||
setInterval(async () => {
|
||||
await lastSlotReceivedMutex.runExclusive(async () => {
|
||||
lastSlotReceived = bulkAccountLoader.getSlot();
|
||||
});
|
||||
lastSlotReceived = bulkAccountLoader.getSlot();
|
||||
}, ORDERBOOK_UPDATE_INTERVAL);
|
||||
|
||||
const userMap = new UserMap(
|
||||
|
||||
Reference in New Issue
Block a user