TypeError: Unknown chunk type 91 while reading record at

[appy-one]

TypeError: Unknown chunk type 91 while reading record at

[neilsutcliffe]

I'm running a next app with the built in api. So single threaded AFAIK.

In terms of how I was using arrays, I was setting the whole array at once, overwriting when updating.

path is like /image/7ddb9f615c2d9947be799b99d17ed59b11c61959

await acebaseFix.ref(path).set(images) // Update via overwrite
await acebaseFix.ref(path).set(newImages) // Create from scratch
await acebaseFix.ref(path).remove() // Delete

Oh, the image array members are structured...

export interface Image {
  width: number;
  height: number;
  img: string;
  original: string;
  url: string;
  cat?: string;
}

I'll give object collections a try, but I think the use case is fine, as I wasn't mutating or changing the array.

Thanks for getting back quickly!

[jkapenieks]

Also, I noticed that my Acebase server script crashed when I was trying to navigate to the node: items/0/3/16766808/item_data on Acebase Web Manager.

[appy-one]

Thanks for reporting this. The best and quickest way for me to investigate this issue is having access to the actual data. If you can send me a copy of the data.db file through wetransfer.com, I will be able to poke into the binary data and see what's wrong and what might have caused it. You can send it to [email protected]

Note that if your data contains personal and/or sensitive data, please create a copy of the db and anonymize your data as much as possible before sending. I will treat your data confidentially and will remove all copies from my system when done!

[appy-one]

Can you please reply

[appy-one]

I'm closing this issue. If you would like this fixed @jkapenieks, please reply. For anyone else running into this issue, please also let me know.

[neilsutcliffe]

Hi. I've ran into this as well. One of my paths refuses to read, which means I can't update or remove it.

Is there anyway to nuke it from orbit without reading it, or to read it and ignore errors

I can send you the database file, but at the moment, I'm just trying to recover my data from the last time I backed up.

[neilsutcliffe]

Tried... await acebaseLocal.ref('/image/7ddb9f615c2d9947be799b99d17ed59b11c61959[1]').set({})

GetTypeError: Unknown chunk type 112 while reading record at "/image/7ddb9f615c2d9947be799b99d17ed59b11c61959[1]" @51,340

I've also tried set([]) to the key directly, and also remove

[appy-one]

@neilsutcliffe Thanks for reporting this.

Are you running AceBase in a multi-process/threaded setup?

Also, I see you are directly updating an array entry, which is discouraged because of reasons described here: Using arrays. In most situations, it's better to use object collections.

I'll add a function to overwrite records today, so you will be able to fix this broken record.

[appy-one]

I've published version 1.22.2,which now supports node recovery. See #142 for more info
Let me know if that works.
Also, if you are sure you are not auto-scaling your app in the cloud, let me know if you find a reproducible way to "break" a node!

[appy-one]

Closing the issue again, feel free to reopen if this is still an issue!

[neilsutcliffe]

Hi. I keep on getting a different issue, but still general corruption. I'm having to back up my database any time I change it. Might look into collections, but I am not doing complex things with arrays...

Either way, I end up corrupting the database when working with Next.js. It could be the hot-reloading is creating multiple connections to the database, which causes this issue.

CorruptRecordError: CORRUPT RECORD: key "img" in "/image/77f9d37ee3c03f1d702a20b11e08cee22210ae5f[0]" (@22,369) refers to address @22,323 which was already used to read "/image/2ffeb7f6c7b5ca0606af6178a7009fb08d69ac4d". Recursive or repeated reading has been prevented.
    at new NodeReader (admin/node_modules/acebase/src/storage-acebase.js:2340:27)
    at loadChildValue (admin/node_modules/acebase/src/storage-acebase.js:2583:40) {
  record: [NodeAddress],
  key: 'img'
}

img in this case is a URL

I've taken to storing different tables in different databases allowing me to revert only one if needed.

[appy-one]

@neilsutcliffe You have to prevent multiple concurrent AceBase instances to the same database to prevent corruption. Reuse 1 single AceBase instance throughout your app to prevent this. Hot reloading should not be an issue, but having 2+ separate instances in your code both altering the same database will corrupt the db.

[neilsutcliffe]

@neilsutcliffe You have to prevent multiple concurrent AceBase instances to the same database to prevent corruption. Reuse 1 single AceBase instance throughout your app to prevent this. Hot reloading should not be an issue, but having 2+ separate instances in your code both altering the same database will corrupt the db.

It's 2 seperate instances with 2 different database files.

However it is reinitialising on hot reloading...

I think I might have to look more into how next.js does this and determines what to recompile. There has to be mechanisms to prevent it from rebuilding.