Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Restoredrill – proves your Postgres backups restore

github.com

Post reply on HN
Print view View original

21–30 of 36 posts

Re: Show HN: Restoredrill – proves your Postgres backups restore

#21

Post by ahmadpiran »

progbits wrote: post #18
ahmadpiran wrote: post #9

Earlier quoted context omitted.

I don't have a good defense here, it's fair. But part of that was because of messy commit history and I wanted to have a clean commit history. Maybe it's not a good reason as well, but it was my real intention and my desire of perfectionism :) reason is that I wanted to take the full control, when ai do the commits you normally don't check the code and you loose the control, well I think it's the most important thing…

Thanks for taking the time to reply! Hope my comment didn't sound too harsh, now that I read it again it might, that's what I get for posting in a hurry. Seems like you put a lot of thought and effort into it which is awesome, so the LLM readme was ruining first impressions. It's a shame but given the explosion in low effort projects it's important to stand out I think.

No worries at all. I take this to improve my working style. I appreciate you circling back to say this :)

Re: Show HN: Restoredrill – proves your Postgres backups restore

#22

Post by ahmadpiran »

sebastiangrill wrote: post #19

The problem i see with this approach is that nearly nobody uses dumps as backups. I consulted Postgres for a few hundred customers and the amount of times that I saw dumps as Backup can probably be counted on two hands. Does not help that the description reads like AI slop.

v1 is pg_dump only because it's the thing that every Postgres install can do natively, and the fastest way to get something really working and tested. Physical backups, pgBackRest specifically are already on the roadmap. One person on interviews doing this kind of testing was on plain dumps and moving off them, about 18 hours to restore versus 2 with a physical backup tool. I don't want to disagree here, I know dump-based isn't a great long term strategy. I will enhance the REAME, thank you for the feedback.

Re: Show HN: Restoredrill – proves your Postgres backups restore

#23

Post by ambicapter »

arjie wrote: post #12
sebmellen wrote: post #5

> There are other tools in this space. Worth naming plainly instead of pretending they don't exist. Why do LLMs like this kind of writing? Why did you need to “plainly” name the competition as opposed to “flamboyantly” naming them? This is a cool idea but the README is absurdly not-to-the-point.

It seems to be a self-instruction. Leak from reasoning token into output token or using output token for reasoning. Perhaps something accidental like constraining number of reasoning tokens leads to model using output tokens for reasoning or perhaps ensuring some minimum user-visible reasoning. Seems like an accidentally introduced artifact. “I should not pretend other implementations don’t exist. I should name them…

That's funny, we talk about the LLM only having a single data channel as a bad thing for security, but I'd never put 2+2 together to realize the LLM can poison itself as well.

Re: Show HN: Restoredrill – proves your Postgres backups restore

#24

Post by csydas »

ahmadpiran wrote: post #20
csydas wrote: post #16

i saw your now dead comment about how no one you interviewed had an answer… just it’s surprising to me that devops folk aren’t in an org that has backup software, cause this kind of testing is standard load out for most backup softwares, and they even solve the space and isolation issue with various instant restore / publishing capabilities where data from the backup is mounted directly to some staging server with po…

One person I talked to had Veeam SureBackup running and still built a separate check for the database. As I know SureBackup proved the VM boots, but it doesn't know or care if your orders table has the right row count, or if a sequence is broken. That needs a tool that actually understand the specific platform or data, like Postgres, not only the underlying VM. Also a lot of people I talked to were never use Veeam or…

aha that’s sort of what i mean though — your concept is two parts: isolated restore and bespoke tests

the isolation part is something that is more difficult due to space and resource constraints, and a lot of backup apps handle that but expect the user to handle the bespoke testing part

your tool covers both but expects the user to provide both which isn’t always feasible for the application teams (not impossible but lot of red tape to request resources and a lot of resistance if the company already has recovery verification tooling)

if i were you i would refocus your pitch on making it work as part of those backup software workflows, “tell your backup team about this test suite and include it in their recovery verification workflows” and i’m sure you’ll get a lot of attention

most of the dba’s probably have a backup and recovery verification solution at their company with the isolation and resources but they don’t have the bespoke application testing

Re: Show HN: Restoredrill – proves your Postgres backups restore

#25

Post by Giefo6ah »

ambicapter wrote: post #23
arjie wrote: post #12

Earlier quoted context omitted.

It seems to be a self-instruction. Leak from reasoning token into output token or using output token for reasoning. Perhaps something accidental like constraining number of reasoning tokens leads to model using output tokens for reasoning or perhaps ensuring some minimum user-visible reasoning. Seems like an accidentally introduced artifact. “I should not pretend other implementations don’t exist. I should name them…

That's funny, we talk about the LLM only having a single data channel as a bad thing for security, but I'd never put 2+2 together to realize the LLM can poison itself as well.

You maybe interested in https://arxiv.org/pdf/2603.12277

Re: Show HN: Restoredrill – proves your Postgres backups restore

#26

Post by ahmadpiran »

csydas wrote: post #24
ahmadpiran wrote: post #20

Earlier quoted context omitted.

One person I talked to had Veeam SureBackup running and still built a separate check for the database. As I know SureBackup proved the VM boots, but it doesn't know or care if your orders table has the right row count, or if a sequence is broken. That needs a tool that actually understand the specific platform or data, like Postgres, not only the underlying VM. Also a lot of people I talked to were never use Veeam or…

aha that’s sort of what i mean though — your concept is two parts: isolated restore and bespoke tests the isolation part is something that is more difficult due to space and resource constraints, and a lot of backup apps handle that but expect the user to handle the bespoke testing part your tool covers both but expects the user to provide both which isn’t always feasible for the application teams (not impossible but…

[dead]

Re: Show HN: Restoredrill – proves your Postgres backups restore

#30

Post by RamblingCTO »

sebmellen wrote: post #5

> There are other tools in this space. Worth naming plainly instead of pretending they don't exist. Why do LLMs like this kind of writing? Why did you need to “plainly” name the competition as opposed to “flamboyantly” naming them? This is a cool idea but the README is absurdly not-to-the-point.

In German it also does "Ich will dich ehrlich kennen lernen" (I want to get to know you honestly/in an honest fashion). Which is super weird because this AI is the opposite of honest. I saw a bio like that on Bumble today ... Super dystopian and weird.
Post reply on HN