Page MenuHomePhabricator

[Identity] Add DDB table for one-time-keys
ClosedPublic

Authored by jon on Jul 31 2023, 7:02 AM.
Tags
None
Referenced Files
F3357348: D8675.id.diff
Sat, Nov 23, 11:37 PM
Unknown Object (File)
Sat, Nov 23, 2:38 PM
Unknown Object (File)
Sat, Nov 23, 2:37 PM
Unknown Object (File)
Sat, Nov 23, 12:55 PM
Unknown Object (File)
Sat, Nov 23, 12:55 PM
Unknown Object (File)
Sat, Nov 23, 4:28 AM
Unknown Object (File)
Tue, Nov 19, 12:55 AM
Unknown Object (File)
Tue, Nov 5, 8:41 AM
Subscribers

Details

Summary

Add new one-time-keys table for both content and notif keys.

https://linear.app/comm/issue/ENG-4535

Depends on D8642

Test Plan
cd services/terraform/dev
./run.sh

The two identity-*-onetime-keys should be created.

The constants in service/identity/ are tested later

Diff Detail

Repository
rCOMM Comm
Lint
No Lint Coverage
Unit
No Test Coverage

Event Timeline

There are a very large number of changes, so older changes are hidden. Show Older Changes

Rebase on combined table work

Correct commit message, rebase on fixes

Not [yet] an expert when it comes to DDB batch writes so leaving this part to @varun

Besides that, LGTM

services/identity/src/database.rs
317–318 ↗(On Diff #29510)

I think this is the same concern as @varun had in https://phab.comm.dev/D8452#250578 - single batch is limited to 25 items and we need to split into chunks - he did this in D8631
Leaving a link to AWS docs: https://docs.aws.amazon.com/amazondynamodb/latest/APIReference/API_BatchWriteItem.html

services/identity/src/ddb_utils.rs
28–29 ↗(On Diff #29510)

This comment is no longer relevant, since we have only one table now

30 ↗(On Diff #29510)

I like this usage of local imports ;)

varun requested changes to this revision.Aug 4 2023, 12:06 PM
varun added inline comments.
services/identity/src/database.rs
317–318 ↗(On Diff #29510)

yep batchWriteItem has limits on size/number of items that we need to adhere to

This revision now requires changes to proceed.Aug 4 2023, 12:06 PM
jon marked 4 inline comments as done.

Address feedback

services/identity/src/ddb_utils.rs
30 ↗(On Diff #29510)

thanks :)

varun requested changes to this revision.Aug 8 2023, 10:19 AM
varun added inline comments.
services/identity/src/database.rs
378

we need to follow the BatchWriteItem limit here, too, right?

This revision now requires changes to proceed.Aug 8 2023, 10:19 AM
jon marked an inline comment as done.

Chunkify other BatchWriteItem

services/identity/src/database.rs
378

yea, my bad.

New devices should only be writing 20 items, and the refreshKeysRequest requests 10. So it shouldn't cause an issue. However, better to be safe than sorry.

varun requested changes to this revision.Aug 9 2023, 3:14 PM

sorry I didn't catch this earlier but we should be consistent about onetime vs one-time/oneTime/OneTime/one_time. the latter matches the white paper, so we should remove all uses of the former. i noted a couple instances where we're using the wrong format inline, but it's not an exhaustive list. some other comments inline too.

also, we have to update the add_user_to_users_table function, right?

services/commtest/tests/identity_onetime_key_tests.rs
18 ↗(On Diff #29632)

identity

35 ↗(On Diff #29632)

fwiw i feel like a full test would also check that we can retrieve the one-time keys as well.

services/identity/src/constants.rs
91 ↗(On Diff #29632)

one-time

services/terraform/modules/shared/dynamodb.tf
257 ↗(On Diff #29632)

i think it should be one-time

258 ↗(On Diff #29632)

one-time

This revision now requires changes to proceed.Aug 9 2023, 3:14 PM
In D8675#259065, @varun wrote:

sorry I didn't catch this earlier but we should be consistent about onetime vs one-time/oneTime/OneTime/one_time. the latter matches the white paper, so we should remove all uses of the former. i noted a couple instances where we're using the wrong format inline, but it's not an exhaustive list. some other comments inline too.

also, we have to update the add_user_to_users_table function, right?

Sigh, I was going off the naming you had in constants.rs. Do you mind if I just create a follow up task to rename all instances of onetime/ONETIME to one-time/ONE_TIME. I really don't want to deal with those kinds of rebase conflicts in the current stack....

jon marked an inline comment as done.
jon added inline comments.
services/commtest/tests/identity_onetime_key_tests.rs
35 ↗(On Diff #29632)

This is done in D8722

services/identity/src/database.rs
1045–1071 ↗(On Diff #29632)

Removing this here means that add_user_... method needs to be updated with adding the otk's to the related table

jon marked an inline comment as done.

"onetime" -> "one-time" for this diff

jon marked 3 inline comments as done.

uggh, not sure why, but I got a stale patch when doing the one-time refactor

jon marked an inline comment as done.

Re-apply feedback which was lost during bad arc patch

varun requested changes to this revision.Aug 16 2023, 6:59 AM

Removing this here means that add_user_... method needs to be updated with adding the otk's to the related table

sorry to keep requesting changes but i'm still not seeing this change in the diff

services/identity/src/constants.rs
93 ↗(On Diff #29632)

are you sure we want to remove this comment?

This revision now requires changes to proceed.Aug 16 2023, 6:59 AM
jon marked an inline comment as done.

Address comments, refactor one time key addtions to single method

Re-introduce comment which was accidentally deleted due to bad rebase

services/identity/src/constants.rs
93 ↗(On Diff #29632)

Another victim of the bad patch.

can you please create a follow up task for renaming affected db methods? e.g. add_user_to_users_table -> add_user_to_ddb, add_password_user_to_users_table -> add_password_user_to_ddb, add_device_to_users_table -> add_device_to_ddb (these are just suggestions, i'm open to other names)

services/identity/src/constants.rs
94–97 ↗(On Diff #30028)

nit on comment location and PARTITION_KEY naming

This revision is now accepted and ready to land.Aug 17 2023, 11:50 AM

Emphasize PARTITION_KEY in comment