none
High login latency RRS feed

  • Question

  • Using my application (with micro-services approach with Kubernetes) I realise it has a problem with long responses. We have 7 micro-services at the moment. On one page we have list of products with multi-tag filter. Each time you click on one of those then there is a a request to one service witch is querying azure postgres with one request. Long story short it is taking the whole process from about 100ms to 2-3s and sometimes even up to 5s to get results back.

    I realise when making simple 'select 1' with psql from cli taking from my laptop 1.5-2s. From inside a container it going down to ~1-1.5s. After login to postgres cli querying is smooth and it's always <100ms often <50ms.

    So my question is what can causing long connection time/authentication?

    Is it possible that this is caused by the low IOP/s? Mine is variable on test (because is Basic VM with 5GB storage) and 100 IOP/s on General Purpose (prod with 5GB storage).

    Please check my additional research on stackoverflow.

    Cannot add link so to to SO domain /questions/57094912/long-connection-time-with-azure-postgres

    Update: Adding Stack Overflow link: https://stackoverflow.com/questions/57094912/long-connection-time-with-azure-postgres




    Thursday, July 25, 2019 9:45 AM

All replies

  • Hi Piotr,

    Per the same comment on the SO thread, are you running postgre on an Azure VM (IaaS) or are you using the Azure Database for PostgreSQL service (PaaS)?

    Thursday, July 25, 2019 5:21 PM
    Moderator
  • This is Azure Database for PostgreSQL service (PaaS)
    Friday, July 26, 2019 8:23 AM
  • Can you send me your instance name and your subscription ID. Please send and email to AzCommunity.

    There is a gateway that listens for requests, and directs traffic to your instance. Please see: Connectivity architecture in Azure Database for PostgreSQL (link)

    I am going to have the product group take a look at your instance to see if this is across the service in the region you are deploy or if this is specific to your deployment. 

    Can you provide some statistics such as timeframe and duration of the login. This will make the investigation more efficient.

    Thank you,

    Mike

    Friday, July 26, 2019 4:31 PM
    Moderator
  • Another quick test is to perform a tracert to see where the delay is at. Please see: How to Use TRACERT to Troubleshoot TCP/IP Problems in Windows (link). 

    Please use the FQDN of you database instance name or the IP Address to use with the tracert utility.

    Thank you,

    Mike

    Monday, July 29, 2019 4:35 AM
    Moderator
  • Want to follow up on this thread to see if a resolution was identified or if you are still experiencing issues?

    Friday, August 9, 2019 5:23 PM
    Moderator