Setting Postgres Timeout for Self-Hosted Solutions

Peter K. · 2025-04-25T18:40:44.811Z

Sorry I keep coming back with more questions! Is there a way to set the postgres timeout on self host? We keep getting timeouts once we have more than like a week of data

Setting Postgres Timeout for Self-Hosted Solutions | Arize AI Community

12 comments

· Sorted by Oldest

RunLLM
APP
·
It seems likely that Phoenix does not explicitly document a way to set the PostgreSQL timeout directly within its configuration. However, you might be able to address timeout issues by adjusting the connection pool settings or the database configuration itself. For instance, increasing the pool_size and max_overflow settings in your database connection configuration might help manage more connections and reduce timeouts, as suggested in a Slack conversation.Would you like more detail on how to configure these settings or manage PostgreSQL connections in Phoenix? Sources: - Slack conversation on connection pool settings
💡 Hint: Mention RunLLM in the thread for followups.
Peter K.
·
Nevermind, we thought it was timeout but we've run out of IDs 🫠 graphql.error.graphql_error.GraphQLError: Int cannot represent non 32-bit signed integer value: 2743948010 Link graphql.error.graphql_error.GraphQLError: Int cannot represent non 32-bit signed integer value: 2743948010
Peter K.
·
I'll open an issue on the github
Roger Y.
·
oh wow
Roger Y.
·
are you able to inspect the network tab to see which query is crashing?

Peter K.

Yep

query ProjectPageQuery(
  $id: GlobalID!
  $timeRange: TimeRange!
) {
  project: node(id: $id) {
    __typename
    ...ProjectPageHeader_stats
    ...StreamToggle_data
    __isNode: __typename
    id
  }
}

fragment ProjectPageHeader_stats on Project {
  traceCount(timeRange: $timeRange)
  tokenCountTotal(timeRange: $timeRange)
  tokenCountPrompt(timeRange: $timeRange)
  tokenCountCompletion(timeRange: $timeRange)
  latencyMsP50: latencyMsQuantile(probability: 0.5, timeRange: $timeRange)
  latencyMsP99: latencyMsQuantile(probability: 0.99, timeRange: $timeRange)
  spanAnnotationNames
  documentEvaluationNames
  id
}

fragment StreamToggle_data on Project {
  streamingLastUpdatedAt
  id
}

Peter K.

{
  "id": "UHJvamVjdDoz",
  "timeRange": {
    "start": "2025-03-29T14:00:00.000Z"
  }
}

Roger Y.
·
ok this looks like token count may be too large and should probably be float instead
Roger Y.
·
thanks for bringing this to our attention. we’ll get it fixed
Peter K.
·
Yep that looks to be the case. Also I had opened this github issue https://github.com/Arize-ai/phoenix/issues/7291 Thank you!
Roger Y.
·
fix has been released in arize-phoenix v8.28.0
Peter K.
·
Awesome, thank you!

RunLLM
APP
·
It seems likely that Phoenix does not explicitly document a way to set the PostgreSQL timeout directly within its configuration. However, you might be able to address timeout issues by adjusting the connection pool settings or the database configuration itself. For instance, increasing the pool_size and max_overflow settings in your database connection configuration might help manage more connections and reduce timeouts, as suggested in a Slack conversation.Would you like more detail on how to configure these settings or manage PostgreSQL connections in Phoenix? Sources: - Slack conversation on connection pool settings
💡 Hint: Mention RunLLM in the thread for followups.
Peter K.
·
Nevermind, we thought it was timeout but we've run out of IDs 🫠 graphql.error.graphql_error.GraphQLError: Int cannot represent non 32-bit signed integer value: 2743948010 Link graphql.error.graphql_error.GraphQLError: Int cannot represent non 32-bit signed integer value: 2743948010
Peter K.
·
I'll open an issue on the github
Roger Y.
·
oh wow
Roger Y.
·
are you able to inspect the network tab to see which query is crashing?

Peter K.

Yep

query ProjectPageQuery(
  $id: GlobalID!
  $timeRange: TimeRange!
) {
  project: node(id: $id) {
    __typename
    ...ProjectPageHeader_stats
    ...StreamToggle_data
    __isNode: __typename
    id
  }
}

fragment ProjectPageHeader_stats on Project {
  traceCount(timeRange: $timeRange)
  tokenCountTotal(timeRange: $timeRange)
  tokenCountPrompt(timeRange: $timeRange)
  tokenCountCompletion(timeRange: $timeRange)
  latencyMsP50: latencyMsQuantile(probability: 0.5, timeRange: $timeRange)
  latencyMsP99: latencyMsQuantile(probability: 0.99, timeRange: $timeRange)
  spanAnnotationNames
  documentEvaluationNames
  id
}

fragment StreamToggle_data on Project {
  streamingLastUpdatedAt
  id
}

Peter K.

{
  "id": "UHJvamVjdDoz",
  "timeRange": {
    "start": "2025-03-29T14:00:00.000Z"
  }
}

Roger Y.
·
ok this looks like token count may be too large and should probably be float instead
Roger Y.
·
thanks for bringing this to our attention. we’ll get it fixed
Peter K.
·
Yep that looks to be the case. Also I had opened this github issue https://github.com/Arize-ai/phoenix/issues/7291 Thank you!
Roger Y.
·
fix has been released in arize-phoenix v8.28.0
Peter K.
·
Awesome, thank you!