S
S
SergioGS2018-11-30 13:58:08
Monitoring
SergioGS, 2018-11-30 13:58:08

How to properly monitor the performance of 1s server in SQL?

Hello!
There is a 1C server with a SQL database.
What exactly to look in SQL Activity Monitor, and the 1C Administration Console to identify the clients that load the server the most? What requests are demanding and what is the server up to at the moment? Users complain about micro freezes in 1s, the conf is self-written, the database weighs 12Gb, SQL has a limit of 19GB RAM, a total of 22 GB RAM on the server, by the way, the server is virtual on Proxmox with SSD. AWE shows that it ate all the memory 5c0117280e441202115888.png5c0117a1c8d0f607737182.png
I want to understand why the friezes? Because of the code? Because of the number of cores or RAM, all the same.

Answer the question

In order to leave comments, you need to log in

2 answer(s)
A
Anton Anton, 2018-11-30
@Fragster

Limit the school to half of the memory, otherwise nothing remains on the 1c server. Well, set up a technological log to track long-running requests and lock conflicts / waits, there is a good visual tool for this (and for analyzing the collected log) devtool1c.ucoz.ru/index/nastrojka_tekhnozhurnala/0-15

E
Evgeniy_27, 2018-12-12
@Evgeniy_27

We need to start by checking the settings:
1) If two roles are on the same server, the setting for working through Shared Memory should be up to
2) run Gilev's tests. General performance calculation and analysis of long queries.
3) set up a memory limit for everyone (sql and 1C). Limit workflows 1C (rphost)
4) I agree with the client above, the technological journal can help, but it will also consume resources, this must be taken into account.
5) Base logs, if they are written to where?
6) Is there TempDB? Where does she lie?
Items 5 and 6 must be on separate physical media. In order not to lose in terms of system performance. But you can skip this point for now and go to monitoring and watch from there.
Now about monitoring:
1) The 1C server needs the frequency of the processor, therefore we monitor the utilization of the processor and the overall load in percent. (1c likes multi-hertz CPUs for multi-threading, you can’t really count on it)
2) queues for disks for reading write ssd is certainly good, but the controller raid may not be enough. Not sure what else is going on there.

Didn't find what you were looking for?

Ask your question

Ask a Question

731 491 924 answers to any question