Redis GET is slow but service log seems good

We are using Redis as cache backend of our Django project and result backend for Celery app. They all use classic sync mode.

The project is hosted on Heroku and use Heroku redis.

From Database overview, I saw the query time of Redis get is slow sometime (this is random but I can see it happen each day)

But from the Heroku redis data dashboard, the Redis service does not have this problem.

The throughput is low so I am really confused.

Since we mainly use Redis as HTML fragment cache backend I am guessing the time also includes the network transfter time and the GET operation return long HTML. But I am not sure and did not have a good way to find more details.

Here is the [ Permalink]

Hello hgyinjian,

Great question! It is a little tough for us to do a side by side comparison of the Heroku redis data dashboard and the Database overview within New Relic, because I don’t have visibility into the Heroku side of things.

That said, there are many variables that could impact the time, for example:

  1. What else redis is doing while these requests are being made.

  2. What else the server that redis lives on is doing at the same time.

  3. You mentioned this as well, network transfer time latency between the server on Heroku’s side and New Relic.

Additionally, I find it helpful to look into how the calculations are being made. Sometimes one products view of things differs from another because of differences in dashboard calculations. In New Relic’s case, we’re looking at an average response time.

I know this may not be the exact answer you are looking for, but I do hope that helps!

1 Like