S3 backup not retrying after error

julien-f

@florent Can you take a look? I wonder if there is a debug option we can enable in the AWS lib.

Andrew

Transient errors (HTTP status codes 400, 408, 500, 502, 503, and 504) 
and throttling errors (HTTP status codes 400, 403, 429, 502, 503, and 509) can all be retried.

So there are more than a few that can be retried.

They also have adaptive and standard retry as well as a configurable number of retries.

olivierlambert

I think we did not have a lot of issues reported on AWS, our concerns are on "similar" implementation but not behaving the same.

florent

@julien-f I didn't find a debug option, but we can put a custom logger. I did quick test with a direct console/log.info and the volume of data logged is huge. I'm trying to make a more reasonable logger

Andrew

@olivierlambert @florent I chose Wasabi because of cost. I understand that non-AWS S3 vendors may not always work as expected but Wasabi is not insignificant.... Local testing to MinIO S3 has never given me any problems.

With tens of thousands of S3 operations I only want to know about the one that failed so I can yell a Wasabi and get them to fix their issue. I have recently been having more nightly issues with them (same retry failures).

So it's a two fold issue. Wasabi is causing problems and XO needs to be more tolerant of problems.

olivierlambert

I'm not saying to use only AWS It's just telling you that we do not have errors reported for AWS customers on S3. It can be for a lot of reasons.

We have invested 10k€ in test hardware to install our own "decent" S3 cluster, so we are taking seriously the fact to try to make it work on other stuff. But as you can see, there's a limit on the number of platforms we can try. Indeed, the solution is to get more debugs, but it's not enjoyable to lose time on potential issues related to 3rd party S3 providers

florent

@Andrew I am sure we'll be able to tune XO to be a little more robust in the near future

florent

@Andrew hi ,

I made a PR that retries 5 times even on error 400, can you test it ?
https://github.com/vatesfr/xen-orchestra/pull/6433

Regards
Florent

fbeauchamp opened this pull request in vatesfr/xen-orchestra

closed fix(s3): retry upload even on error 400 #6433

Andrew

@florent Thanks. I'm running it. I'll report after a few days.

Andrew

@florent Last night's failure (commit 81ae8)...

{
      "data": {
        "type": "VM",
        "id": "f80fdf51-65e5-132d-bb2a-936bbd2814fc"
      },
      "id": "1663912365483:2",
      "message": "backup VM",
      "start": 1663912365483,
      "status": "failure",
      "tasks": [
        {
          "id": "1663912365570",
          "message": "clean-vm",
          "start": 1663912365570,
          "status": "failure",
          "end": 1663912403372,
          "result": {
            "name": "InternalError",
            "$fault": "client",
            "$metadata": {
              "httpStatusCode": 500,
              "extendedRequestId": "jOYV90/W5XHJFnOq1mlfpaMT/T9EV4/EnSluEni+p9TJQykrtI0cJMntJqFThy/PvX/LN0XX4xXS",
              "attempts": 3,
              "totalRetryDelay": 369
            },
            "Code": "InternalError",
            "Detail": "None:UnexpectedError",
            "RequestId": "85780FD1B7DFCB7C",
            "HostId": "jOYV90/W5XHJFnOq1mlfpaMT/T9EV4/EnSluEni+p9TJQykrtI0cJMntJqFThy/PvX/LN0XX4xXS",
            "message": "We encountered an internal error.  Please retry the operation again later.",
            "stack": "InternalError: We encountered an internal error.  Please retry the operation again later.\n    at throwDefaultError (/opt/xo/xo-builds/xen-orchestra-202209221033/node_modules/@aws-sdk/smithy-client/dist-cjs/default-error-handler.js:8:22)\n    at deserializeAws_restXmlGetObjectCommandError (/opt/xo/xo-builds/xen-orchestra-202209221033/node_modules/@aws-sdk/client-s3/dist-cjs/protocols/Aws_restXml.js:4356:51)\n    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)\n    at async /opt/xo/xo-builds/xen-orchestra-202209221033/node_modules/@aws-sdk/middleware-serde/dist-cjs/deserializerMiddleware.js:7:24\n    at async /opt/xo/xo-builds/xen-orchestra-202209221033/node_modules/@aws-sdk/middleware-signing/dist-cjs/middleware.js:11:20\n    at async StandardRetryStrategy.retry (/opt/xo/xo-builds/xen-orchestra-202209221033/node_modules/@aws-sdk/middleware-retry/dist-cjs/StandardRetryStrategy.js:51:46)\n    at async /opt/xo/xo-builds/xen-orchestra-202209221033/node_modules/@aws-sdk/middleware-flexible-checksums/dist-cjs/flexibleChecksumsMiddleware.js:56:20\n    at async /opt/xo/xo-builds/xen-orchestra-202209221033/node_modules/@aws-sdk/middleware-logger/dist-cjs/loggerMiddleware.js:6:22\n    at async S3Handler._createReadStream (/opt/xo/xo-builds/xen-orchestra-202209221033/@xen-orchestra/fs/dist/s3.js:261:15)\n    at async S3Handler.readFile (/opt/xo/xo-builds/xen-orchestra-202209221033/@xen-orchestra/fs/dist/abstract.js:326:18)"
          }
        },
        {
          "id": "1663912517635",
          "message": "snapshot",
          "start": 1663912517635,
          "status": "success",
          "end": 1663912520335,
          "result": "85b00101-5704-c847-8c91-8806195154b4"
        },
        {
          "data": {
            "id": "db9ad0a8-bce6-4a2b-b9fd-5c4cecf059c4",
            "isFull": false,
            "type": "remote"
          },
          "id": "1663912520336",
          "message": "export",
          "start": 1663912520336,
          "status": "success",
          "tasks": [
            {
              "id": "1663912520634",
              "message": "transfer",
              "start": 1663912520634,
              "status": "success",
              "end": 1663912549741,
              "result": {
                "size": 251742720
              }
            },
            {
              "id": "1663912551469",
              "message": "clean-vm",
              "start": 1663912551469,
              "status": "success",
              "end": 1663912629752,
              "result": {
                "merge": false
              }
            }
          ],
          "end": 1663912629752
        }
      ],
      "end": 1663912629752
    },

florent

@Andrew Have you got any retrying writing file message in the logs ?

Andrew

@florent No, I did not see that in the logs. I did see this problem is bigger than I thought.

It happens more often than just causing a VM backup failure. It happens during the merge or other checks which causes the backup process to destroy (remove) parts of other VM backups.

 Clean VM directory 
 parent VHD is missing
 parent VHD is missing
 parent VHD is missing
 some VHDs linked to the backup are missing
 some VHDs linked to the backup are missing
 some VHDs linked to the backup are missing
 some VHDs linked to the backup are missing

and

 Clean VM directory 
 VHD check error
 some VHDs linked to the backup are missing