ServiceException: (504)

Support
1

Hi,

When trying to update embeddings from haystack, I am getting the following error


Reason: Gateway Timeout
HTTP response headers: HTTPHeaderDict({'content-type': 'text/plain', 'content-length': '24', 'date': 'Wed, 07 Jun 2023 23:47:51 GMT', 'server': 'envoy'})
HTTP response body: upstream request timeout

I have been running the same code snippet for the couple of weeks but it has stopped working for me today. I was wandering if I hit a free tier limit somewhere.

I would be grateful if anyone can help me solve the problem

2

Hi @ashutosh. Please check your email, I just sent you a proactive ticket on this issue. The short version is your index has a very high degree of cardinality which is preventing it from being built.

3

hey haven’t received any mail yet, can you retry?

4

Could you please post the solution here so we could all benefit from it??
Everything was working fine and I have a demo tomorrow to many potential customers and this happens!! Need urgent help!!

5

Also getting this error, was working fine yesterday.
I have a much smaller index (1/100th) the size that works ok, but my main large index has been off and on all day with this same error.

6

It decided to work fine again now.
That’s good, but unreliable. We still have no clue why it went down last night and when that would happen again.

7

A post was split to a new topic: 504 errors and cannot connect in the web console

8

Hello all,

504 error is a genric error and there can be a number of different causes, some trigger by user actions and other times there is an internal error on our side.

Unfortunely there is not a sliver bullet soultion and each case has to be looked at in turn.

In most case the problem is when the index is overloaded for exampling:

In both of these cases increasing the size of the pod resolves the issue.

The good news is that our new serverless index address these points and many others.

9

@pwozny and @wissam.moussa if you have not already please do open a support case and we can look into these issues.

10

Thanks! We’re trying out the new serverless index this week