HPCC-19494 New IBYTI logic not working properly #11043
Merged
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
|
https://track.hpccsystems.com/browse/HPCC-19494 |
richardkchapman
force-pushed
the
ibyti-fix
branch
3 times, most recently
from
04607fa
to
3aaa983
Compare
Code was assuming meaning in subChannels array that was not actually valid. Make the assumption valid by allocating subChannels in a more cyclic fashion. Signed-off-by: Richard Chapman <rchapman@hpccsystems.com>
richardkchapman
force-pushed
the
ibyti-fix
branch
from
a060db7
to
92e2189
Compare
ghalliday
reviewed
Apr 19, 2018
roxie/ccd/ccd.hpp
Outdated
| @@ -192,6 +193,12 @@ class RoxiePacketHeader | |||
| return SUBCHANNEL_MASK << (SUBCHANNEL_BITS * subChannel); | |||
| } | |||
|
|
|||
| unsigned getRespondingSubChannel() const // NOTE - 0 based | |||
| { | |||
| unsigned bitpos = ffs(retries); | |||
There was a problem hiding this comment.
Need to define ffs for windows
roxie/ccd/ccdqueue.cpp
Outdated
| * for any given channel. | ||
| * | ||
| * To determine which subchannel is the "primary" for a given query packet, a hash value of fields from the packet header | ||
| * is used, modulo the number of subchannels for on this channel. The slave on this subchannel will respond immediately. |
roxie/ccd/ccd.hpp
Outdated
| @@ -96,6 +96,7 @@ extern unsigned myNodeIndex; | |||
|
|
|||
| #define SUBCHANNEL_MASK 3 | |||
| #define SUBCHANNEL_BITS 2 // allows for up to 7-way redundancy in a 16-bit short retries flag, high bits used for indicators/flags | |||
| #define MAX_SUBCHANNEL 7 | |||
There was a problem hiding this comment.
Add comment // (16-2) / SUBCHANNEL_BITS - or calculate
roxie/ccd/ccdqueue.cpp
Outdated
|
|
||
| void init(unsigned channel) | ||
| { | ||
| mySubChannel = subChannels[channel]-1; |
There was a problem hiding this comment.
Could do with comments on the channels/subChannels variables to indicate exactly what they mean.
roxie/ccd/ccdqueue.cpp
Outdated
| return (liveSubChannels[channel] & mask) == mask; | ||
| } | ||
| public: | ||
| unsigned getIbytiDelay(unsigned subChannel) const // NOTE - zero-based |
There was a problem hiding this comment.
Clearer if the parameter was called primarySubChannel, same for the function below.
roxie/ccd/ccdqueue.cpp
Outdated
| channels[channelCount++] = channel; | ||
| } | ||
| for (int i = 0; i < channelCount; i++) | ||
| subChannelInfo[i].init(i); |
There was a problem hiding this comment.
subChannelInfo[channels[i]].init(channels[i]);
or merge into the previous loop
roxie/ccd/ccdqueue.cpp
Outdated
| while (primarySubChannel != mySubChannel) | ||
| { | ||
| unsigned channelMask = SUBCHANNEL_MASK << (SUBCHANNEL_BITS * primarySubChannel); | ||
| if ((h.retries & ROXIE_RETRIES_MASK) == channelMask) | ||
| if (primarySubChannel == theirSubChannel) |
There was a problem hiding this comment.
still trying to understand this change...
There was a problem hiding this comment.
It was supposed to make this code clearer! Net effect will be the same.
|
Automated Smoketest: ✅ Unit tests result:
Regression test result:
HPCC Stop: OK
|
|
@mckellyln looks good to me and ready for testing. |
Reduce IBYTI delays more gradually - treating a single timed-out IBYTI as indicative of a dead node seems to cause far more failed IBYTIs than prior logic did. This version should be cleaner and thus easier to understand and fix in the future, should cope with more than 2 subchannels, and should behave identically to the pre-6.4.12 code when there are only 2 subchannels. Signed-off-by: Richard Chapman <rchapman@hpccsystems.com>
richardkchapman
force-pushed
the
ibyti-fix
branch
from
a82238e
to
c002832
Compare
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.
Code was assuming meaning in subChannels array that was not actually valid.
Make the assumption valid by allocating subChannels in a more cyclic fashion.
Signed-off-by: Richard Chapman rchapman@hpccsystems.com
Type of change:
Checklist:
Testing:
Debugged through and observed values and code routes being more what I was expecting.
Note that I don't know if this is the ONLY issue with the new IBYTI code, nor does it explain the memory usage issue which suggests that a Roxie with a high rate of ineffective IBYTI is leaking - this needs to be investigated elsewhere.