Closed Bug 919026 Opened 11 years ago Closed 9 years ago

switch /report/index to using postgres

Categories

(Socorro Graveyard :: Middleware, defect)

x86
macOS
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: rhelmer, Unassigned)

References

Details

We have raw crashes in postgres now, and will have the processed dump soon as well (bug 907305). Let's switch /report/index off of hbase and onto postgres only - the only places it will still be needed are: 1) priority processing (e.g. the crash was throttled) 2) downloading minidump (logged-in users only) This means that the basic /report/index page will work even when hbase is down.
report/index/ loads both ProcessedCrash and RawCrash. But the lesser traffic to HBase the better. In other words, we can make this bug about just switching out the backed being used for RawCrash.
Is there a way to completely eliminate Hbase from the equation? Since we store the raw crash in the database, and we'll soon store the processed crash in the database, theoretically we could eliminate Hbase entirely. That could/would make it easier to eliminate Hbase as a dependency for Socorro altogether.
Assignee: nobody → bsavage
By "eliminate Hbase as a dependency" I mean it would be possible to easily stand up a Socorro that excludes Hbase (not that we would eliminate it entirely from our setup)
(In reply to Brandon Savage [:brandon] from comment #3) > By "eliminate Hbase as a dependency" I mean it would be possible to easily > stand up a Socorro that excludes Hbase (not that we would eliminate it > entirely from our setup) Yes, this is already the case - just use filesystem for raw storage. In fact this is the checked-in default in the configs right now.
(In reply to Robert Helmer [:rhelmer] from comment #4) > (In reply to Brandon Savage [:brandon] from comment #3) > > By "eliminate Hbase as a dependency" I mean it would be possible to easily > > stand up a Socorro that excludes Hbase (not that we would eliminate it > > entirely from our setup) > > Yes, this is already the case - just use filesystem for raw storage. In fact > this is the checked-in default in the configs right now. This is what you get if you follow http://socorro.readthedocs.org/en/latest/installation.html too.
Fixing this bug would make it possible to use Postgres for all raw JSON access I think (but not raw minidumps), and I think we should do that for the Mozilla implementation. We'll continue to store raw/processed JSON and the minidumps in HBase though (the JSON for analytic purposes and the latter for archival purposes) It means that everything except for processing incoming crashes and downloading minidumps will work on crash-stats.mozilla.com when HBase is down or otherwise unreachable.
Depends on: 921125
Assignee: bsavage → nobody
Nope. We're using S3 now instead.
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
Product: Socorro → Socorro Graveyard
You need to log in before you can comment on or make changes to this bug.