Page MenuHomePhabricator

[native] Blocklist for backed up tables
ClosedPublic

Authored by michal on Mar 29 2024, 6:47 AM.
Tags
None
Referenced Files
Unknown Object (File)
Sun, Nov 24, 12:41 AM
Unknown Object (File)
Sun, Nov 3, 8:03 AM
Unknown Object (File)
Oct 12 2024, 7:07 PM
Unknown Object (File)
Oct 12 2024, 7:07 PM
Unknown Object (File)
Oct 12 2024, 7:07 PM
Unknown Object (File)
Oct 12 2024, 7:07 PM
Unknown Object (File)
Sep 12 2024, 1:55 AM
Unknown Object (File)
Aug 29 2024, 7:50 PM

Details

Summary

ENG-6918 : Improve backup handling of new tables

Currently tracking if a table is contained in backup happens in two places. Compactions use a SQL query that iterates over tables and runs DELETE FROM on backup file. Logs use a vector of table list to watch for. This diff unifies them and has a single blocklist, which informs both compaction and log creation.

Test Plan
  • Created a draft, created a compaction, created another draft (this one is only captured in logs)
  • Re-logged, restored backup, made sure both drafts are there
  • Added "drafts" table to backedUpTableBlocklist and repeated the test. This time neither of the logs existed after restore

Diff Detail

Repository
rCOMM Comm
Lint
No Lint Coverage
Unit
No Test Coverage

Event Timeline

native/cpp/CommonCpp/DatabaseManagers/SQLiteQueryExecutor.cpp
41–48

There are a few differences:

  • integrity_store is blocked, as we don't need it for backup
  • persist_storage is blocked, all O(1) data in redux persist that we will need, will be stored in synced_metadata

Also this doesn't block communities table which means that it will now be included in logs gathering (so it was a bug because this table was included in a compaction but not in logs).

1116

Wasn't sure where to put this, but at this point we have run the migrations so we should have a static db schema (for the lifetime of SQLiteQueryExecutor).

ashoat added inline comments.
native/cpp/CommonCpp/DatabaseManagers/SQLiteQueryExecutor.cpp
41–48
native/cpp/CommonCpp/DatabaseManagers/SQLiteQueryExecutor.cpp
41–48

Could you explain what you mean by compaction?

Currently working on sqlite stores for AuxUser (will store farcaster ids of mutuals) and ThreadActivity. Do either of these need to go on the blocklist? Possibly thread activity?

https://linear.app/comm/issue/ENG-5310/move-threadactivitystore-to-sqlite
https://linear.app/comm/issue/ENG-7405/auxiliary-user-store-follow-ups

marcin requested changes to this revision.Apr 2 2024, 3:03 AM

LGTM for me but please stick to the convention to use SQLiteStatementWrapper.

native/cpp/CommonCpp/DatabaseManagers/SQLiteQueryExecutor.cpp
1090

The convention is not to use sqlite3_stmt directly but to use SQLiteStatementWrapper class that handles freeing memory and error handling for you.

This revision now requires changes to proceed.Apr 2 2024, 3:03 AM
native/cpp/CommonCpp/DatabaseManagers/SQLiteQueryExecutor.cpp
1082

could you put this below and group it with other functions hidden behind the ifdef to avoid adding more macros to this file?

1090

agree with that

native/cpp/CommonCpp/DatabaseManagers/SQLiteQueryExecutor.h
32

isn't this more readable now?

Move function, rename, use SQLiteStatementWrapper

native/cpp/CommonCpp/DatabaseManagers/SQLiteQueryExecutor.cpp
41–48

Backup consists of two parts:

  • Compaction which is a SQLite dump
  • Backup logs, which we capture when there are new changes to the sqlite db (so we don't have to upload a large compaction everytime and only smaller deltas)

I agree, thread activity should be shouldn't be backed up.

This revision is now accepted and ready to land.Apr 4 2024, 8:13 AM