Re: Intermittent missing chunk ERRORS

On 4/26/25 2:24 PM, Laurenz Albe wrote:
> On Sat, 2025-04-26 at 11:09 -0600, Sbob wrote:
>> I am working with a client, they have an app that is running into error's like this:
>>
>> SQL Error [XX001]: ERROR: missing chunk number 0 for toast value 44915477 in pg_toast_5321896
>>
>> I thought it was a corruption / REINDEX issue, however I have run REINDEXon the index,
>> REINDEX on the table and even a reindexdb on the database. Today they passed me another
>> error (the one above) and once I was able to login I ran the same query the app team
>> referenced and it ran without any errors. I asked them to re-run the query and they
>> informed me that it now works for them as well.
>>
>> I assume that if this were a corrupt index issue then the query would continue to fail
>> every time.
>>
>> Has anyone seen this type of intermittent index error before? and thoughts on where to
>> look /how to debug this?
> The error is data corruption.
>
> If it is transient, that could mean a couple of things:
>
> - a TOAST index was corrupted, and REINDEX fixed it
>
> - faulty RAM corrupted data, but as soon as the corrupted buffer was evicted,
>    the problem was gone
>
> - somebody updated or deleted the row, and the corruption is hidden
>
> Yours,
> Laurenz Albe


I did REINDEX and the error came back, however this last time when they 
informed me about the error they sent the select statement throwing the 
error and once I logged in I ran their select statement and it worked, 
no REINDEX had been run, and then I asked them to run it again and it 
worked for them as well

So, the first option above is not the answer and I dont think anyone 
updated the row, their process is mostly reads but I will verify, so 
maybe the RAM issue? the servers are cloud VM's in a private cloud