Page MenuHomePhabricator

[services] Backup - generate backup id
ClosedPublic

Authored by karol on May 13 2022, 2:26 AM.
Tags
None
Referenced Files
F2829552: D4028.diff
Fri, Sep 27, 4:53 PM
Unknown Object (File)
Sun, Sep 15, 9:43 AM
Unknown Object (File)
Sun, Sep 15, 9:43 AM
Unknown Object (File)
Sun, Sep 15, 9:43 AM
Unknown Object (File)
Sun, Sep 15, 9:43 AM
Unknown Object (File)
Sun, Sep 15, 9:43 AM
Unknown Object (File)
Fri, Sep 13, 12:40 AM
Unknown Object (File)
Fri, Sep 6, 1:42 PM

Details

Summary
Test Plan

Run the blob service
Run the backup service
Use grpc-playground to make a call creating a new backup

Either log out the generated backup id or check in the database

Diff Detail

Repository
rCOMM Comm
Lint
No Lint Coverage
Unit
No Test Coverage

Event Timeline

Harbormaster returned this revision to the author for changes because remote builds failed.May 13 2022, 2:26 AM
Harbormaster failed remote builds in B9079: Diff 12625!
karol edited the test plan for this revision. (Show Details)
karol added a reviewer: tomek.

services build passed

tomek requested changes to this revision.May 17 2022, 3:19 AM
tomek added inline comments.
services/backup/src/Reactors/server/CreateNewBackupReactor.cpp
12–15 ↗(On Diff #12625)

We're not using userID so why do we have it as a parameter?

What's the point of having both UUID and timestamp when timestamp is a part of UUID?

This revision now requires changes to proceed.May 17 2022, 3:19 AM
services/backup/src/Reactors/server/CreateNewBackupReactor.cpp
12–15 ↗(On Diff #12625)

Good point, I'll remove it.

Ok, if so, we can have backupID == generateUUID
Only one thing: can you provide where'd you get this info from? Thanks.

tomek added a reviewer: ashoat.

Adding @ashoat because choosing id schema is an important decision

services/backup/src/Reactors/server/CreateNewBackupReactor.cpp
12–15 ↗(On Diff #12625)

It is described in a couple of places, e.g. https://en.wikipedia.org/wiki/Universally_unique_identifier#Format

But in the future, please try to research it yourself and verify if the proposition is valid.

services/lib/src/GlobalTools.h
9 ↗(On Diff #12836)

After this change we no longer need this in this diff. It is needed though later in the stack.

karol added inline comments.
services/backup/src/Reactors/server/CreateNewBackupReactor.cpp
12–15 ↗(On Diff #12625)

I was looking in the boost docs as we use boost and couldn't find it. Essentially, we use boost's code for this so maybe a more solid source would be to provide something from the docs/code? Nevertheless, I suppose we can assume that the term UUID means the same for Wikipedia's article author and boost authors.

But in the future, please try to research it yourself and verify if the proposition is valid.

Why? I implemented some code, and you proposed this change. So effectively, you proposed some code here. Shouldn't that be you in this case who backs up their idea?

services/lib/src/GlobalTools.h
9 ↗(On Diff #12836)

Sure, I can move this.

services/backup/src/Reactors/server/CreateNewBackupReactor.cpp
12–15 ↗(On Diff #12625)

Why? I implemented some code, and you proposed this change. So effectively, you proposed some code here. Shouldn't that be you in this case who backs up their idea?

Think about this in this way: your diff is a proposition of solution. I have some doubts about it. So you should prove that my doubts are invalid. So my idea was the result of doubts I have about your idea.

remove ID_SEPARATOR from here

ashoat requested changes to this revision.May 19 2022, 10:09 PM

I'm open to this idea, but I think there's been a slight misinterpretation of my Linear comment here. I was suggesting that we use a UUID for the device ID, but it would be the same UUID for the device (persisted in Redux, for instance). And then the backup would be <device_id>:<timestamp>, where the device ID is a UUID, and the timestamp is the current timestamp.

I think Karol interpreted my comment as suggesting that we use a UUID for the whole backup in place of the device ID. So it's basically <backup_uuid>:<timestamp>.

I think @palys-swm's suggestion to remove the timestamp makes sense for Karol's interpretation, but not for mine.

One potential advantage of mine might be that it might be easier to group backups by device ID (in a batch job, for instance), and another is that it may be possible to force an efficient index that shows you all of the backups for a given device (not sure if DynamoDB supports this).

What do you all think?

This revision now requires changes to proceed.May 19 2022, 10:09 PM

I don't mind and I can implement any solution, really. So in the case you're describing, the device_id is just something we're going to get from the client, right?

ashoat requested changes to this revision.May 20 2022, 7:07 AM

Yeah, that's right!

This revision now requires changes to proceed.May 20 2022, 7:07 AM

I added a diff before this one that reads the device id from the client, so we can use it here D4099

tomek added inline comments.
services/backup/src/Reactors/server/CreateNewBackupReactor.cpp
12

I know we call this method only from one place where we know that deviceID is set, but maybe we can make it safer by checking if deviceID is set?

54

We should remove this comment, right?

This revision is now accepted and ready to land.May 23 2022, 10:49 AM
services/backup/src/Reactors/server/CreateNewBackupReactor.cpp
12

ok

54

right