Dump task fails with "Cannot allocate memory (os error 12)"

[JamieDixon]

Hi there,

We're trying to upgrade our production Meilisearch instance but have been unable to create a dump of the current data. The message we're returned from the task is "Cannot allocate memory (os error 12)". We've tried to increase the available memory on the server itself but this didn't have any effect.

Our Meilisearch version states:

{
    "commitSha": "unknown",
    "commitDate": "unknown",
    "pkgVersion": "1.0.2"
}

Here is the full task response:

{
    "uid": 597664,
    "indexUid": null,
    "status": "failed",
    "type": "dumpCreation",
    "canceledBy": null,
    "details": {
        "dumpUid": null
    },
    "error": {
        "message": "Cannot allocate memory (os error 12)",
        "code": "internal",
        "type": "internal",
        "link": "https://docs.meilisearch.com/errors#internal"
    },
    "duration": "PT3.796615598S",
    "enqueuedAt": "2023-10-23T11:02:25.065555256Z",
    "startedAt": "2023-10-23T11:02:25.066579548Z",
    "finishedAt": "2023-10-23T11:02:28.863195146Z"
}

[curquiza]

Hello @JamieDixon

Can you try with v1.4.0?

Also, if it does not work, can you give us more details to know how you trigger the error?
What is the size of your machine (RAM and CPU)? What is the size of your dataset?

[JamieDixon]

@curquiza I notice that if I delete the folder that contains my locally-created index, meilisearch throws some errors, which indicates that a reference to the index must be stored somewhere else as well as the folders themselves.

Does this mean I need to tell meilisearch to re-scan the indexes folder in order for it to create references to the existing indexes?

(Clutching at straws here just in case this is relevant)

[JamieDixon]

@curquiza Small progress: I re sshd the data.ms folder again and now when I run meilisearch, I get all the same errors as I do on Digital Ocean. I'm assuming this means that all the indexes are present and I simply don't have enough memory to run many of the request.

This takes us back to step 1 and the options look like either 1) upgrade the available memory in Digital Ocean or 2) write a script to re-index all of our data into meili cloud.

Thanks for all your help today. It's made this impossible task more bearable to have someone in the mud with me 👍

[curquiza]

Hello @JamieDixon
Good to know you found a kind of solution.
Your situation was indeed really weird 🤔 I hope increasing the size of the machine will help, at least for creating a dump manipulate your data as you wish.

[JamieDixon]

@curquiza Hey. I tried increasing the memory on the server and it had no useful effect. In the end I decided to sign up to meilisearch cloud and I wrote a script to create the indexes and add the data. In total we only have 186,157 documents so I'm going to chalk this one up to an issue with the server and move on.

We'll be better off on the cloud product I think anyway and maintaining the infrastructure was always an unnecessary burden.

Thanks again for all your help. I'll close this question off for now.

[curquiza]

Thank you @JamieDixon for letting us know. Hope you will appreciate the Cloud experience 😊

Also, if you have more questions we organize a webinar on November 2nd about helping our users make a transition into the Cloud experience. You can register here: https://us06web.zoom.us/meeting/register/tZMlc-mqrjIsH912-HTRe-AaT-pp41bDe81a#/registration

Have a nice day! 👋