1

We’ve just migrated from Sitecore 8.2.5 on-prem to 9.3 Azure PaaS. Performance seems ok most of the time, but we’re seeing regular spikes to 100% DTU on the web db.

The db is standard S2 50 DTU and we’re running 8 websites. I’m not convinced this adequate but need to convince my company.

Is there any optimization that we can do? And what tier are others using for their db’s?

Thanks.

7
  • 1
    kb.sitecore.net/articles/582735
    – Mark Cassidy
    Commented Dec 8, 2020 at 21:13
  • The one solution there - increase the DTU from 50 to 100 as a start point Commented Dec 9, 2020 at 5:22
  • Is there a rhythm in the spikes? You should try to determine what is happening when those spikes occur.. is anything else spiking as well (cpu, ...)?
    – Gatogordo
    Commented Dec 16, 2020 at 13:50
  • @Madgibbon We have a similar question answered below please check if that helps sitecore.stackexchange.com/questions/23126/… Commented Dec 16, 2020 at 19:28
  • We identified some of the spikes are coming from client data clean up job. Two of our sites make use of virtual users, this was cause the client data table to grow by 4GB per day. The clean up agent runs every 4 hours for 10 mins. However after 10 mins the job failed due to the amount of data and rolled back. This was causing 20 mins of 100% DTU every four hours.We have now disabled the Sitecore job and setup and Elastic Job to clean the table using a stored procedure at 01:00 every day.
    – Madgibbon
    Commented Dec 17, 2020 at 19:50

1 Answer 1

0

Could be load, could be unoptimized queries. I would suggest increasing to S3 to test. If ongoing cost is a concern, you can also try grouping multiple DBs together in an elastic pool to share DTUs.

1
  • Did consider Elastic Pools, but the Sitecore Azure documentation recommends against it.
    – Madgibbon
    Commented Dec 17, 2020 at 19:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.