11-21-2022
This week 202211210928 - CHASE THE FUTURE 202211211009
Lunch options:
- Chaat Bhavan
- The Press Cafe Delivery & Takeout | 1987 Santa Rita Road Pleasanton | Menu & Prices | DoorDash
- Ikes, Curry Cravings, both in the same plaza
Outcomes
Summary
- Decided to use The Archive for all my notes, and I feel really good about it! I’m already comfortable with using this app, and it’s a lot better if I impose my own structure in a free-form way than use the Obsidian template, even if it does work really well for certain people.
- Going to hop back on the Dominion vibe!
- Message for Daniel for 202212090801:
- I understand that different people have different expectations, and load balancing isn’t an exact science. At the same time, I know that I’m the best person to advocate for protecting my own focus time, and I want to make sure my time is being treated at the same level as everyone else’s.
Artifacts
- Left interview feedback — https://workflowy.com/#/7f732e1b7c7e
- Decided lunch options — see above
- Did two code reviews:
- [maindb resilience] implement resilient composite connection creation by KeunwooLee-at · Pull Request #51279 · Hyperbase/hyperbase
- [maindb resilience] Use dedicated mysql connection setting for fallback replica by JunyiWang-at · Pull Request #49649 · Hyperbase/hyperbase
- [maindb resilience] Enable error injection for CONSISTENT_READ_REPLICA_FALLBACK strategy by JunyiWang-at · Pull Request #51215 · Hyperbase/hyperbase
- Talked to Bailey about interview time.
- incident reflections 202211220922
- app flags don’t work well on the worker child? Why are they so delayed
- although, I looked at the logs and seems like the admin flag is getting correctly populated through the local disk loader. I have no idea what’s happening here.
- do we need some way of force polling the config loader?
- it keeps trying to retry connections, and it seems like in between retries, it does not repoll the admin flag. If it keeps retrying and timing out, then that means that the admin flag change won’t take place.
- this also relates to event loop blockage.
- how are we still seeing staleness at this point!! https://opensearch-applogs.shadowbox.cloud/_dashboards/app/discover#/doc/airtable-applogs-index/.ds-airtable-applogs-fe226b8fc6b0a573f9bcc8730e25cc6c-000019?id=D-NYnYQBwhIOKlh9D3Ij - I think I understand this now
- why is the temporary thing not error, but the withmainconnection does error! Is it because it runs in a new domain every time?
- even if this hides the error, the fact that it prevents errors is useful.
- this could happen due to event loop blockages and connection timeouts! We should add a fix in withMainConnectionAsync to account for this. In general, we probably shouldn’t have production critical code that accounts for this in such a simplistic way.
- how does the admin flag refresh process work in the worker child?
- do we not log properly when the main connection idles out?
- investigate these errors:
- “Retrying mysql connection open after error”
- “Communication with MySQL appears to be hung while connecting”
- if we crash on errors
- https://opensearch-applogs.shadowbox.cloud/_dashboards/goto/ca33ca58775d990304836994cce6a6d2?security_tenant=global — connection hung, on db shards cache domain.
- we’re running with retries. if something is retrying (but it’s stuck, because of timeouts), we could have multiple things trying to fetch at the same time. Add more logging in db shards cache, or, if there is a promise already being awaited, then make the new request also wait for the promise.
- app flags don’t work well on the worker child? Why are they so delayed
Notes:
- I don’t like Arc. It feels a bit more sluggish, and the process of creating a new tab feels inherently slower for some reason than Chrome.
- I feel like part of the value-add of Arc is that it makes tabs easier to manage, but the way it does that is just by making it harder to make new tabs.
- I like the notes and easel features, but I haven’t had a reason to use them.
uid: 202211210801 tags: #journal