1271466 - Sync load testing based on production traffic patterns and data

Status: RESOLVED DUPLICATE of bug 1459326

(Cloud Services Graveyard :: Server: Sync, defect)

Description

[Ryan Tilder [:rtilder]]

We'd like to do some DB performance comparisons based off some actual production traffic patterns and data.  Anonymized and fuzzed data, obviously, but since the data is already encrypted and user IDs are strictly numeric I don't think it will be much of a problem.

The goal is to have one or more scripts that can take a tcpdump data file and generate a data set that can be readily played back in either traditional or batched mode.

Operations has provided a suitable packet dump.  Progress with processing TCP headers and packet flows with Python has proved full of unexpected entertainment.

Comment 1

[Ryan Kelly [:rfkelly]]

FWIW, I think this will be useful enough in general, that we should put the script in the server-syncstorage repository alongside the other management scripts.

Comment 2

[Karl Thiessen [:kthiessen, he/him]]

This seems to be dormant for now; I'll pick it back up if/when it becomes a P1/P2.

Comment 3

[Ryan Kelly [:rfkelly]]

Removing this as a blocker for the batch-upload work

Comment 4

[Ryan Kelly [:rfkelly]]

Rebecca, do you have a different tracking bug for your recent work on this topic?  Maybe we can close this one as a dupe.

Comment 5

[Rebecca Billings [:rbillings]]

Yes, this has been reanimated in bug 1459326. Closing.