The Wireshark support is for a third party tool. Do we need to do a full test plan?
If so, this is going to be a huge effort requiring many more developers. The support for LNet is over 1,000 lines of code, and the Lustre support is over 12,000 lines of code. Further, without adequate internal Lustre documentation, there is no easy way to ensure this support is correct without involving a lot of the current Lustre experts.
Previous thinking was to just get the old Wireshark support working (it did not work at all) at a very basic level, and then correct what is probably many mistakes in the protocol interpretation over time. The ticket LU-1434 just got the basics working. There are known bugs in the Lustre interpretation.
The Wireshark support is for a third party tool. Do we need to do a full test plan?
If so, this is going to be a huge effort requiring many more developers. The support for LNet is over 1,000 lines of code, and the Lustre support is over 12,000 lines of code. Further, without adequate internal Lustre documentation, there is no easy way to ensure this support is correct without involving a lot of the current Lustre experts.
Previous thinking was to just get the old Wireshark support working (it did not work at all) at a very basic level, and then correct what is probably many mistakes in the protocol interpretation over time. The ticket
LU-1434just got the basics working. There are known bugs in the Lustre interpretation.