Thread: Database log in
All,
I am experiencing intermitently, slow, log in times. My connection scenario:
- Kube pod is hosted in Ohio, in AWS account A.
- Using psql, a log in attempt is initiated from the pod.
- RDS Postgres Instance is hosted in Ohio, in account B. host has 8 vCPU and 32 GB of RAM. CPU and IO are not a bottle neck.
- We experience intermittent log in slowness.
My questions are around what happens at log in time; what are the sequence of events? Also, how much “stress” does log in have to the DBMS? How many concurrent logins can the DBMS handle in a short time frame? Can 40 sessions be initiated within 90 seconds? 60 seconds?
Please, I would appreciate debugging suggestions.
Thank you in advance, Kim.
My questions are around what happens at log in time; what are the sequence of events? Also, how much “stress” does log in have to the DBMS?
How many concurrent logins can the DBMS handle in a short time frame? Can 40 sessions be initiated within 90 seconds? 60 seconds?
Please, I would appreciate debugging suggestions.
David G. Johnston wrote on 1/24/2024 4:34 PM:
On Wednesday, January 24, 2024, Heckler, Kim M <kim.heckler@nationwide.com> wrote:My questions are around what happens at log in time; what are the sequence of events? Also, how much “stress” does log in have to the DBMS?
A decent amount.How many concurrent logins can the DBMS handle in a short time frame? Can 40 sessions be initiated within 90 seconds? 60 seconds?
It depends.Please, I would appreciate debugging suggestions.
Don’t? At this point you’ve probably hit the point where connection pools are the best way forward. Especially since 8 is much less than 40.David J.
MichaelDBA <MichaelDBA@sqlexec.com> writes: > I don't have a problem with hundreds or thousands of logins per minute > on RDS or bare metal, even with only 8 CPUs and 32GB RAM. There is > something else in play here... Yeah. The OP didn't mention the authentication method in use, but I could believe some sort of performance glitch if it consults an external server (for Kerberos, AD, LDAP, etc). Or maybe it's basically a network problem between client and server. regards, tom lane
>> MichaelDBA <MichaelDBA@sqlexec.com> writes: >> > I don't have a problem with hundreds or thousands of logins per minute >> > on RDS or bare metal, even with only 8 CPUs and 32GB RAM. There is >> > something else in play here... >> >> Yeah. The OP didn't mention the authentication method in use, but I could believe some sort of >> performance glitch if it consults an external server (for Kerberos, AD, LDAP, etc). >> Or maybe it's basically a network problem between client and server. >> >> regards, tom lane >> Auth method is local password. The AWS RDS Instance does have Kerberos enable so there is ONE role that authenticates via AD. I was thinking, maybe, that the local ID was mistakenly trying to authenticate via AD, failing, and then authenticating locally?? I do have a ticket open with AWS to inquire about this. Also, I will use strace to gather more info. Thank you, Kim.