# Corrupted database

**URL:** <https://discuss.dgraph.io/t/corrupted-database/18852>\
**Category:** Dgraph\
**Created:** [August 17, 2023, 12:12pm UTC](https://discuss.dgraph.io/t/corrupted-database/18852 "2023-08-17T12:12:03Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![magentarepeater](https://avatars.discourse-cdn.com/v4/letter/m/6a8cbe/32.png) [@magentarepeater](https://discuss.dgraph.io/u/magentarepeater)\
**Post date:** [August 17, 2023, 12:12pm UTC](https://discuss.dgraph.io/t/corrupted-database/18852/1 "2023-08-17T12:12:04Z")

</div>

### What version of Dgraph are you using?

v21.03

### What is the hardware spec (RAM, OS)?

AWS EC2 m6a.4xlarge, Linux

### What happened?

The database was suddenly unavailable. After restart of the whole cluster (docker-compose) I got these logs

```auto
graphAlpha_1 | W0817 11:37:36.004553 17 mvcc.go:148] Error Key with size 171798 exceeded 65000 limit. Key:
graphAlpha_1 | 00000000 00 00 00 00 00 00 00 00 00 00 05 74 69 74 6c 65 |...........title|
graphAlpha_1 | 00000010 02 01 67 66 67 67 68 64 66 67 68 64 69 66 67 6a |..gfgghdfghdifgj|
graphAlpha_1 | 00000020 64 68 66 67 6a 6b 64 68 67 6b 6a 68 66 6a 6b 67 |dhfgjkdhgkjhfjkg|
graphAlpha_1 | 00000030 68 6a 64 6b 66 6b 67 6a 68 64 66 6a 67 68 6a 6b |hjdkfkgjhdfjghjk|
graphAlpha_1 | 00000040 6b 6a 6b 66 67 68 6b 64 6a 66 67 6b 6a 64 66 67 |kjkfghkdjfgkjdfg|
graphAlpha_1 | 00000050 68 6b 6a 64 66 68 67 6a 6b 64 68 66 67 6a 6b 68 |hkjdfhgjkdhfgjkh|
graphAlpha_1 | 00000060 64 66 6a 67 6b 68 64 6a 66 67 68 6a 6b 64 66 68 |dfjgkhdjfghjkdfh|
...

```

Continues to print a lot of hex code.

This happened in production. Starting the cluster with a new volume works. We needed to go back to a backup.

We’re trying to find out what the problem is. Any ideas? Thank you.

---

<div class="post-metadata">

**Author:** ![amanmangal](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/amanmangal/32/1249_2.png) [@amanmangal](https://discuss.dgraph.io/u/amanmangal)\
**Post date:** [August 18, 2023, 8:30am UTC](https://discuss.dgraph.io/t/corrupted-database/18852/2 "2023-08-18T08:30:32Z")

</div>

This issue seems related to this fix [fix(mutation): validate mutation before applying it by mangalaman93 · Pull Request #8623 · dgraph-io/dgraph · GitHub](https://github.com/dgraph-io/dgraph/pull/8623). Essentially, a mutation has a much larger value and which leads to a much larger index key. This PR disallows such mutations.

---

<div class="post-metadata">

**Author:** ![magentarepeater](https://avatars.discourse-cdn.com/v4/letter/m/6a8cbe/32.png) [@magentarepeater](https://discuss.dgraph.io/u/magentarepeater)\
**Post date:** [August 21, 2023, 12:17pm UTC](https://discuss.dgraph.io/t/corrupted-database/18852/3 "2023-08-21T12:17:22Z")

</div>

Thanks so much for your response. That makes a lot of sense. We’ll update to v23 asap.

Just to get a better understanding: There was a value (so not the key itself) that exceeded a certain limit? And second: Could seeing the word “title” in the error logs be an indicator that it was the “title” key?

Thanks!

---

<div class="post-metadata">

**Author:** ![amanmangal](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/amanmangal/32/1249_2.png) [@amanmangal](https://discuss.dgraph.io/u/amanmangal)\
**Post date:** [August 23, 2023, 3:19pm UTC](https://discuss.dgraph.io/t/corrupted-database/18852/4 "2023-08-23T15:19:14Z")

</div>

First: The value in the data becomes the key for the index which eventually leads to the problem. And for second, that seems like a reasonable guess. Predicates are usually in the beginning of the key.