Herbert Wolverson
|
a4c8093401
|
Circuit data now includes TCP retransmits. Made the tables a little less unpleasant to look at.
|
2024-03-19 14:13:04 -05:00 |
|
Herbert Wolverson
|
b01ab42a3b
|
Nice highlighting for top 10 buttons.
|
2024-03-19 13:29:04 -05:00 |
|
Herbert Wolverson
|
e43bd00717
|
Fix typo
|
2024-03-19 12:47:56 -05:00 |
|
Herbert Wolverson
|
7e6146b79a
|
Add the protocols top 10 summary
|
2024-03-19 12:43:49 -05:00 |
|
Herbert Wolverson
|
c6feebe229
|
Add plumbing for the top 10 displays. Display ethertypes from expired flows.
|
2024-03-19 12:29:52 -05:00 |
|
Herbert Wolverson
|
b69748dcf6
|
REmember to scale the unit on display
|
2024-03-19 09:01:12 -05:00 |
|
Herbert Wolverson
|
a2afa7fa9d
|
This should fix the country RTTs.
|
2024-03-19 08:48:05 -05:00 |
|
Herbert Wolverson
|
e45bedc7ce
|
One more try before I go and eat something...
|
2024-03-18 16:13:06 -05:00 |
|
Herbert Wolverson
|
ebc3ab73da
|
Make the country RTTs actually add up
|
2024-03-18 16:05:37 -05:00 |
|
Herbert Wolverson
|
019d301913
|
Make the country RTTs actually add up
|
2024-03-18 16:04:43 -05:00 |
|
Herbert Wolverson
|
bad4429729
|
Last commit, but now with sensible logic.
|
2024-03-18 15:57:07 -05:00 |
|
Herbert Wolverson
|
98155a04c3
|
Only update flow RTT if there is data for it.
|
2024-03-18 15:49:31 -05:00 |
|
Herbert Wolverson
|
7b6da8e32e
|
Limit RTT tracking to flows that have achieved 1 Mbps or more.
|
2024-03-18 10:19:50 -05:00 |
|
Herbert Wolverson
|
dec1440b31
|
Flow data: the RTT tracker now keeps track of if new data has arrived for a flow, and only reports an RTT if it has. This prevents stuck flows with no new values coming in from biasing our results.
|
2024-03-16 09:16:05 -05:00 |
|
Herbert Wolverson
|
c9600f831d
|
RTT gathering improvement: RTTs are grouped per circuit, and then added as a single value to help preserve per-circuit sample stability.
|
2024-03-16 09:08:31 -05:00 |
|
Herbert Wolverson
|
b6eb08751c
|
Fix order of magnitude on the display histogram.
|
2024-03-15 13:27:12 -05:00 |
|
Herbert Wolverson
|
d0e575f317
|
Add compatibility with the 5.x kernel version eBPF verifier.
|
2024-03-15 13:24:58 -05:00 |
|
Herbert Wolverson
|
6ce7e7a3f9
|
Strongly type RTT data in the throughput tracker to reduce confusion.
|
2024-03-15 13:03:49 -05:00 |
|
Herbert Wolverson
|
b8937bf9c2
|
Basic web UI update to show numbers in the right units
|
2024-03-15 12:55:23 -05:00 |
|
Herbert Wolverson
|
9d0e536089
|
Use a dual-structure to store RTT in both directions.
|
2024-03-15 12:29:47 -05:00 |
|
Herbert Wolverson
|
9938a94caf
|
Comment out console log that was annoying me
|
2024-03-15 12:19:59 -05:00 |
|
Herbert Wolverson
|
beda42194a
|
Include RTT events per second tracking in the /api/stats call
|
2024-03-15 12:18:19 -05:00 |
|
Herbert Wolverson
|
f44af37670
|
WIP - Compiles RTT data into the tracker, strong type RTTs to clarify the unit confusion. Web side is not done yet.
|
2024-03-15 12:15:11 -05:00 |
|
Herbert Wolverson
|
56b170f7e4
|
Truly minimal RTT to userspace eBPF ringbuffer implementation. This
breaks a lot of things - no RTTs are currently recorded or acted
upon. The goal is to measure the scale of the load hit for receiving
these events.
|
2024-03-15 09:45:49 -05:00 |
|
Herbert Wolverson
|
3ad023f69c
|
Fix circuit flows display from last patch.
|
2024-03-15 09:02:07 -05:00 |
|
Herbert Wolverson
|
318e20dad7
|
These numbers line up on the iZones tracker, with wireshark dumps for comparison. This is a WIP - don't apply this. Unbreaking some of my previous code.
|
2024-03-15 08:32:06 -05:00 |
|
Herbert Wolverson
|
56dd4b6750
|
It's not quite done yet, but I spent the last 2 hours matching observed RTT, checking with Wireshark, and looking to see if the numbers line up. Implements a small buffer for RTT values per flow. Shrinks some stack entries. Will require a map rebuild.
|
2024-03-14 14:07:06 -05:00 |
|
Herbert Wolverson
|
69508d2753
|
Remove two variables and use pointers instead to save stack space.
|
2024-03-14 12:40:33 -05:00 |
|
Herbert Wolverson
|
a937820a87
|
Move rate estimate from 64-bits to 32-bits, saving some bytes.
|
2024-03-14 11:03:43 -05:00 |
|
Herbert Wolverson
|
8ec361b095
|
Fix a reporting message that was meant to be compile-conditional.
|
2024-03-14 10:42:22 -05:00 |
|
Herbert Wolverson
|
fbb3960b50
|
Warning fix
|
2024-03-14 08:52:36 -05:00 |
|
Herbert Wolverson
|
eab9df4874
|
Missing file from previous commit
|
2024-03-13 14:56:59 -05:00 |
|
Herbert Wolverson
|
fb91e8313a
|
Very silly commit - adds a /showoff page to the node manager firing particles from all endpoints at my ISP in Missouri. Will turn into something useful in the future.
|
2024-03-13 14:45:53 -05:00 |
|
Herbert Wolverson
|
a63ff0a6f1
|
Rename all flowbee "retries" to "tcp_retransmits" in code, and
"retransmits" in visible HTML.
|
2024-03-13 08:20:36 -05:00 |
|
Herbert Wolverson
|
fae96280b9
|
Actually limit the top 10 endpoints to being 10
|
2024-03-12 16:41:57 -05:00 |
|
Herbert Wolverson
|
c16f06b0ab
|
Sort countries by download bytes not upload
|
2024-03-12 15:47:29 -05:00 |
|
Herbert Wolverson
|
66a19c04a4
|
First example of an actual summary report - we can group flow endpoints in a 5-minute report.
|
2024-03-12 15:19:07 -05:00 |
|
Herbert Wolverson
|
1fb5838ebe
|
Remove a test call that was spamming the logs.
|
2024-03-12 14:46:47 -05:00 |
|
Herbert Wolverson
|
e46aafe5ae
|
Silly performance: making 12k mutex locks when I need one was really dumb.
|
2024-03-12 14:20:35 -05:00 |
|
Herbert Wolverson
|
5a3f90412d
|
Working geocode system, albeit not as useful as I hoped.
|
2024-03-12 14:02:36 -05:00 |
|
Herbert Wolverson
|
55f24cf71b
|
Revert previous bad idea
|
2024-03-12 11:38:19 -05:00 |
|
Herbert Wolverson
|
07239b3d24
|
That should avoid some flow duplication that was hitting too hard.
|
2024-03-12 11:30:13 -05:00 |
|
Herbert Wolverson
|
3ca7ca8a0d
|
Lower threshold for capturing RTT by rate
|
2024-03-12 10:52:41 -05:00 |
|
Herbert Wolverson
|
e20d6d39b0
|
Another try at a flow system that cleans up.
|
2024-03-12 10:29:08 -05:00 |
|
Herbert Wolverson
|
3d9b52e627
|
More cleanup and logic improvements based on observation. In particular, flow cleanup is more accurate now, and we're a bit more aggressive in what RTT data we accept.
|
2024-03-12 08:57:29 -05:00 |
|
Herbert Wolverson
|
213a27498e
|
Improve the flow cleanup logic.
|
2024-03-12 08:21:33 -05:00 |
|
Herbert Wolverson
|
79247e07f0
|
Fix a really silly mistake. I did NOT mean to always store the worst RTT we've ever seen.
|
2024-03-11 15:26:23 -05:00 |
|
Herbert Wolverson
|
d93726e538
|
Only report RTT for flows exceeding 4kbps - to eliminate noise from basically idle connections.
|
2024-03-11 15:03:47 -05:00 |
|
Herbert Wolverson
|
71fd1d558f
|
Probable fix for building this on earlier kernel versions.
|
2024-03-11 14:04:41 -05:00 |
|
Herbert Wolverson
|
445cdcda81
|
Remove most of the Heimdall mode 1 path, cleaning up the execution path now that we have global flow tracking.
|
2024-03-11 13:27:42 -05:00 |
|