Page Menu
Home
ClusterLabs Projects
Search
Configure Global Search
Log In
Files
F4525352
No One
Temporary
Actions
View File
Edit File
Delete File
View Transforms
Subscribe
Mute Notifications
Flag For Later
Award Token
Size
3 KB
Referenced Files
None
Subscribers
None
View Options
diff --git a/TODO b/TODO
index 82e168c1..7274a56c 100644
--- a/TODO
+++ b/TODO
@@ -1,115 +1,108 @@
0.1
---
compress/encrypt: review API crypto->nsscrypto + compress->zlibcompress and ABI
current feels clumpsy and we can save memcpy around
in case we don't encrypt/compress
compress must happen before encrypt, should we compress all? what's
the price for uncompress on already uncompressed buffer? do we need
a bit to indicate if pckt is compressed or not?
is it worth compressing pings? do they compress at all?
config bits:
- should crypto and compress be part of handle_new? or new api calls?
we won't allow runtime changes, so i don't think there is much
benefit to have it outside of knet_handle_net for now.
We maybe able to spare some duplication between knet_h and knet_cfg struct
link layer:
-one we have pckt inspector in place, we need a function to:
+we need a function to:
+void dst_cache_links_update(knet_host *host);
+that will order link_entries by priority/weight for RR.
-int get_dst_links(knet_host *host, int *fd_array, size_t fd_array_size);
-
-that will call into dst_cache
-
-dst_cache is populated by link status changes, cache calculation should
-probably be in its own thread. this is a bit touchy to implement, specially
-due to rwlocks required to update the cache. Look into link status changes
-notification (maybe a signal to a thread? or write to a pipe)
-
-if all links for a host are down, then clear circular_buffers (this
-is necessary to detect restarts and a bunch of other things)
+Look into link status changes notification (maybe a signal to a thread? or write to a pipe)
* add priority/weight to link config struct
-* add active/active to link config struct
+
+we might need to update the cache on link removal...
ring:
* support 8 listener per node.
* index pointer list
general:
* what to log
log to file: who logins, config changes and link status changes
log to vty: cmd execution failure
* consider splitting tests/* based on what they test (tap_test -> libtap/) ?
vty:
* split cmd_files
* basic show info
* fix check_param for ip/prefix/crypto
0.2
---
libtap:
* add man pages
* improve tests to cover thread safety and more error codes
0.3
---
* benchmark tests:
- all critical paths in ring.c
0.4
---
general:
* add statistics
* add init script / spec file
* v4/v6 (bindv6only check via cli)
vty:
* add optional options
* sorting
* tab completion on options
0.5
---
* pong count
0.6
---
* dynip
0.7
---
* review ring api for libknetring shared lib
* review tap api for libknettap shared lib
* review vty api for libknetvty shared lib
1.0-pre
-------
tests:
* coverity
* unit test:
- test all public APIs
- write ad-doc tests for internal complex functions
- not required to test return codes from external libs directly
(we are not testing glibc)
- test all code paths we write
- no 0.9/1.0 release without max testing coverage for the core
File Metadata
Details
Attached
Mime Type
text/x-diff
Expires
Thu, Jun 26, 6:46 PM (1 d, 8 h)
Storage Engine
blob
Storage Format
Raw Data
Storage Handle
1959427
Default Alt Text
(3 KB)
Attached To
Mode
rK kronosnet
Attached
Detach File
Event Timeline
Log In to Comment