packetblaster: use default number of tx/rx descriptors #883
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Adopt the value for num_descriptors that is recommended in the optimization guidelines: https://github.com/snabbco/snabb/blob/next/src/doc/performance-tuning.md
This depends on the default number of descriptors being sane. See merge 192f00c to ensure this. This seems to work fine - tested line rate with minimum packet size on 10x10G per packetblaster process: https://twitter.com/lukego/status/721693707251105792
lukego
added a commit
to lukego/snabb
that referenced
this pull request
Apr 19, 2016
Enable a really important assertion that the packet freelist does not overflow when we allocate a new batch of packets. This should not be a performance issue: the check is on a slow-path where we are allocating a thousand new packets. This is actually very important. I found a bug that packetblaster would crash with heap corruption / segfault when you run it with more than 3 network interfaces at the same time. This is because it allocated too many packets for transmit/receive buffers and that error was not caught unless "developer debug" mode was enabled. (That problem is separately alleviated by snabbco#883 that reduces the descriptor queues from 32K packets to 2K packets each.)
Packets are allocated dynamically on demand and this is only an artifical limit of "more than anybody should ever need" for catching unbounded allocations (leaks). (The limit also serves to make the freelist simple i.e. a fixed-size array.) The previous value of 100,000 is too low to serve its purpose: it is reasonable to allocate more than 100,000 packet buffers when managing many network receive queues and wanting to avoid packet drops. See discussion at: snabbco#882 (comment)
Needed to make the test suite pass. See related: snabbco#882 (comment)
Needed to work around CI problems.
|
Had to manually kill off SnabbBot for d616abf because the nfvconfig selftest got stuck. Can't really tell why this happened. |
eugeneia
added a commit
that referenced
this pull request
Apr 26, 2016
# Conflicts: # src/program/packetblaster/packetblaster.lua
takikawa
pushed a commit
to takikawa/snabb
that referenced
this pull request
Aug 5, 2017
Distinguish state vs config grammars
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This depends on the default number of descriptors being sane and for that reason this branch includes a merge of #882.
This seems to work fine - tested line rate with minimum packet size on 10x10G per packetblaster process: https://twitter.com/lukego/status/721693707251105792