Files
Florent BEAUCHAMP 14e4c7d62c fix(backups): timeout NBD transfer omnibus fixes (#10306)
* fix(backups): timeout was not used correctly for NBD transfers
* fix(nbd-client): evict a dead client in multi instead
actual code was retrying 5 times, each time with a 60s timeout
multi also have its own retry on conneciton
since the client used is determnistic for a block, any stalled client
will be hammerred for a lot of time before giving up

this PR evict a nbd client that is failing to read a block (with its
retry logic)

* fix(nbd-client/multi): better spread read accross client
the actual code derivate the clientId from the block index
we have no information on the block distribution, only hopes that
it's random enough
this PR introduce a mechanism to use a round robin to select
the next client

* fix: correct fall back to full with /without nbd on qcow2
2026-08-28 10:57:17 +02:00

170 lines
5.4 KiB
JavaScript

import { asyncEach } from '@vates/async-each'
import { NBD_DEFAULT_BLOCK_SIZE } from './constants.mjs'
import NbdClient from './index.mjs'
import { createLogger } from '@xen-orchestra/log'
const { warn } = createLogger('vates:nbd-client:multi')
export default class MultiNbdClient {
#clients = []
#nbdConcurrency
#nextClient = 0
#options
#readAhead
#settings
get exportSize() {
return this.#clients[0].exportSize
}
constructor(settings, { nbdConcurrency = 8, readAhead = 16, ...options } = {}) {
this.#readAhead = readAhead
this.#options = options
this.#nbdConcurrency = nbdConcurrency
if (!Array.isArray(settings)) {
settings = [settings]
}
this.#settings = settings
}
/**
*
* open nbdConcurrency connections to NBD servers
* it must obtain at least one connection to succeed
* it tries to spread connections on multiple host
*
* @returns {Promise<void>}
*/
async connect() {
const candidates = [...this.#settings]
const baseOptions = this.#options
const _connect = async () => {
if (candidates.length === 0) {
return
}
// a little bit of randomization to spread the load
const nbdInfo = candidates[Math.floor(Math.random() * candidates.length)]
const client = new NbdClient(nbdInfo, {
...baseOptions,
readAhead: Math.ceil(this.#readAhead / this.#nbdConcurrency),
})
try {
await client.connect()
this.#clients.push(client)
} catch (err) {
client.disconnect().catch(() => {})
// do not hammer unreachable hosts, once failed, remove from the list
const candidateIndex = candidates.findIndex(({ address }) => address === nbdInfo.address)
if (candidateIndex >= 0) {
// this candidate may have already been deleted by another parallel promise
candidates.splice(candidateIndex, 1)
}
warn(`can't connect to one nbd client`, { err })
// retry with another candidate (if available)
return _connect()
}
}
// don't connect in parallel since this can lead to race condition
// on distributed systems ( like the NBD server of the XAPI)
for (let i = 0; i < this.#nbdConcurrency; i++) {
await _connect()
}
if (this.#clients.length === 0) {
const error = new Error(`Fail to connect to any Nbd client`, { nbdInfos: this.#settings })
error.code = 'NO_NBD_AVAILABLE'
throw error
}
if (this.#clients.length < this.#nbdConcurrency) {
warn(
`incomplete connection by multi Nbd, only ${this.#clients.length} over ${this.#nbdConcurrency} expected clients`
)
}
}
/**
* @returns {Promise<void>}
*/
async disconnect() {
await asyncEach(this.#clients, client => client.disconnect(), {
stopOnError: false,
})
}
/**
*
* @param {number} index
* @param {number} size
* @returns {Promise<Buffer>}
*/
async readBlock(index, size = NBD_DEFAULT_BLOCK_SIZE) {
const clientId = this.#nextClient++ % this.#clients.length
const client = this.#clients[clientId]
try {
return await client.readBlock(index, size)
} catch (err) {
// client.readBlock() already exhausted its own retries/reconnects: this connection is dead.
// Evict it so future reads stop being routed to it, and retry this read on a surviving
// client, since the data is usually still reachable through the others.
this.#evict(client)
if (this.#clients.length === 0) {
throw err
}
warn(`evicted a dead nbd client, retrying block ${index} on a remaining client`, { err })
return this.readBlock(index, size)
}
}
// no-op if `client` was already evicted by a concurrent failed read on the same client
#evict(client) {
const i = this.#clients.indexOf(client)
if (i === -1) {
return
}
this.#clients.splice(i, 1)
client.disconnect().catch(() => {})
}
/**
*
* @param {AsyncGenerator<Buffer>} indexGenerator
*/
async *readBlocks(indexGenerator) {
// default : read all blocks
const readAhead = []
const makeReadBlockPromise = (index, size) => {
const promise = this.readBlock(index, size)
// error is handled during unshift
promise.catch(() => {})
return promise
}
// read all blocks, but try to keep readAheadMaxLength promise waiting ahead
for (const { index, size } of indexGenerator()) {
// stack readAheadMaxLength promises before starting to handle the results
if (readAhead.length === this.#readAhead) {
// any error will stop reading blocks
yield readAhead.shift()
}
readAhead.push(makeReadBlockPromise(index, size))
}
while (readAhead.length > 0) {
yield readAhead.shift()
}
}
/**
* returns the map of the file with holes, zeros and data, useful to handle efficiently sparse source *
*
* @returns {Promise<{ offset: number, length: number, type: number }[]>}
* A promise that resolves to an array where each object represents a segment:
* - `offset` — The byte offset from the start.
* - `length` — The size of the segment in bytes.
* - `type` — A numeric code indicating the segment type (0 means no data).
*/
async getMap(signal) {
// ask the map from one of the connected client
return this.#clients[Math.floor(this.#clients.length * Math.random())].getMap(signal)
}
}