Tom.f on Thu, 13 Jul 2017 08:53:23


I am expreincing very very high latency with postgres.

I have a VM running on Azure West Europe and the DB set up in West Europe as well.

On the VM I have confluence (that was running fine with low latency to the DB in a different VM but I want to migrate to Azure).

Since starting the evaluation I saw that confluence resposne times are very bad so I checked their performance dashboard and saw the latency to the DB (which is tested by doing the following simple query according to Atlassian docs: select * from CLUSTERSAFETY) is between 5-20ms and spikes to up to 80-170ms every few quries. According to their docs it should be around 0-1ms while 2-5ms indicate a latency problem.

I also took the same query and tried it with my computer (in Israel) and got 300-400ms per response.

I can kind of understand why from Israel it takes so much time but can you explain why it takes so much time in your own region?



Saloni Sonpal on Sat, 15 Jul 2017 00:29:42

Hi Tom.f,

Thank you for reporting this. We would be happy to look into this, but I need a few more details from you to dig deeper into the issue. Can you please share your server name and timestamp when you experienced the latency? Also, we know that the local VM connection latency is going to be lower compared to remote managed database server, but for comparison can you please share the VM numbers too? Additionally, do you have info on the particular queries that spike up the times?

Feel free to email me at saloni(dot)sonpal(at)microsoft(dot)com



Shipsheep on Tue, 01 Aug 2017 14:53:19

Same issue here, latency is unacceptable even for the simplest SELECT statements with no other load on the DB.

Yes, I know that 50 / 100 compute units is the low end of the future price scale, but this is a dev instance of an existing app that has been happily with the PostgreSQL DB hosted on a 1 CPU / 2 GB RAM VM ... also in EU West.

I already shared the details of the DB instance with Sunil in the context of

bjomi on Tue, 08 Aug 2017 13:28:35

I experience the same thing. Basically it is totally unusable as it is now. And I want to know if it is supposed to be like that. My intention was to run a live site (Django) on it. Is that doable or should I use another solution (running a da in a VM) or some other cloud provider. I can't see any reason for releasing a service that does not practically work (performance wise) so that is why I'm asking.

I posted about my issues here before:

Best regards,