Is there some feature to make Crawlee continue crawl from the stop place?

[qkxie]

I am new to Crawee. I have add 10000 urls to the RequestQueue. After the crawler finished about 5000 urls, my computer restarted and the crawler stopped.

So how can I continue to crawl the left urls instead of retry all urls?

Yes I know I can do this work by third party components such as Redis just like what I do in Scrapy or other crawler frames.

My question is that can I do this job just by Crawee itself?

[B4nan]

The crawler state should be still there (in the storage folder) and should be respected next time you run the crawler, all you need to do is disable the auto-purging of the state, e.g. via CRAWLEE_PURGE_ON_START env var, e.g. CRAWLEE_PURGE_ON_START=0 npm start.

[qkxie]

Thanks, I have made a try and it indeed continue from the stop point.

[Yash-946]

can you share how to do it like what file you create and what you written

[ylwu]

@Yash-946 just run CRAWLEE_PURGE_ON_START=0 npm start.

[jawspeak]

If you want to have it retry errored out requests in the request queue, the way I've found to do it is by parsing the request_queues/default/*.json files and archiving (renaming) ones with failures and then edit some attributes and starting up the crawl without passing in any new requests. (The queue is resumed).

My goal was to run until failures, then resume where it failed, giving more attempts after changing something. This assumes it fails the _MAX_RETRIES quantity and if it fails less than that and you interrupt/terminate the crawl, then it will not detect those to resume. There may be another better way, hopefully.

// Call this with some --resume flag codepath or something
export function _resumeRetryFixRequestQueue(crawleeQueueDir: string) {
    Configuration.getGlobalConfig().set('purgeOnStart', false); // let it resume
    const archivedTime = toFileSafeName(nowEastern()); // just makes a name
    const files = fs.readdirSync(crawleeQueueDir);
    files.forEach(file => {
        const filePath = path.join(crawleeQueueDir, file);
        if (file == ".DS_Store") { // no need for these, delete
            fs.rmSync(filePath);
            return;
        }
        if (path.extname(file) != ".json") {
            console.warn(`Unexpected non-Json file found: ${file}, ignoring`);
            return;
        }

        const reqFromQueue = readJsonFile<_CrawleeInternalRequest>(filePath);
        // Look at each request and if the retryCount is our max, then archive and edit to retry.
        // We might be able to better decide if it was errored out by looking at the __crawlee.state if it is 5 (ERROR_HANDLER), for now use retryCount.
        if (reqFromQueue.retryCount >= _MAX_RETRIES) {
            console.log(`File: ${file} max retries, will reset to retry again, key: ${reqFromQueue.uniqueKey}, full file: ${JSON.stringify(reqFromQueue)}`);
            const archivePath = path.join(crawleeQueueDir,
                'archived_' + archivedTime,
                renameFileBeforeExtension(file, `_archived_${archivedTime}`)); // just rename it so it is ignored but we have a record of the original file
            fs.cpSync(filePath, archivePath);

            // Manipulate the state to enable it to resume
            reqFromQueue.retryCount = 0;
            reqFromQueue.orderNo = Date.now();
            const jsonData = JSON.parse(reqFromQueue.json);
            jsonData['retryCount'] = 0;
            jsonData['handledAt'] = null;
            jsonData['userData']['__crawlee'].state = RequestState.REQUEST_HANDLER; // should be 3, from 5 during errors (ERROR_HANDLER)
            reqFromQueue.json = JSON.stringify(jsonData);
            writeJsonFile(filePath, reqFromQueue);
        }
    });
}


// Copied from request-queue.ts in crawlee to deserialize the json files in the request queue.
type _CrawleeInternalRequest = {
    id: string
    orderNo: number | null
    url: string
    uniqueKey: string
    method: string
    retryCount: number
    json: string // deserialize manually and edit pinpoint values
}