New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 688412 link

Starred by 1 user

Issue metadata

Status: Archived
Owner:
Closed: Feb 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 2
Type: Bug



Sign in to add a comment

chromeos2-devserver5 has 4 ethernet NICs but only 2 have traffic

Project Member Reported by semenzato@chromium.org, Feb 3 2017

Issue description

From "sar -n DEV" :

05:45:01 AM     bond0   6701.68  15625.15   1136.34  22156.77      0.00      0.00      2.76      9.08
05:45:01 AM       em2   5350.33   3813.15    472.50   5201.93      0.00      0.00      1.63      4.26
05:45:01 AM       em4      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
05:45:01 AM        lo     29.29     29.29      4.75      4.75      0.00      0.00      0.00      0.00
05:45:01 AM       em1   1351.35  11812.00    663.84  16954.84      0.00      0.00      1.13     13.89
05:45:01 AM       em3      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
05:50:01 AM     bond0   9214.28  10582.17   5640.43  14874.79      0.00      0.00      2.76      6.09
05:50:01 AM       em2   7461.17   1913.66   5509.44   2295.67      0.00      0.00      1.63      4.51
05:50:01 AM       em4      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
05:50:01 AM        lo     35.83     35.83     51.22     51.22      0.00      0.00      0.00      0.00
05:50:01 AM       em1   1753.10   8668.51    130.98  12579.12      0.00      0.00      1.13     10.30
05:50:01 AM       em3      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
05:55:01 AM     bond0  18474.90  19678.63  14356.92  27189.52      0.00      0.00      2.76     11.14
05:55:01 AM       em2   4203.24   6253.15    335.37   7874.04      0.00      0.00      1.64      6.45
05:55:01 AM       em4      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
05:55:01 AM        lo     28.18     28.18      3.23      3.23      0.00      0.00      0.00      0.00
05:55:01 AM       em1  14271.66  13425.47  14021.55  19315.48      0.00      0.00      1.13     15.82
05:55:01 AM       em3      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
06:05:01 AM     bond0  15017.92  15968.84  10741.84  22258.02      0.00      0.00      2.76      9.12
06:05:01 AM       em2   8663.06   8483.68   4215.52  11460.69      0.00      0.00      1.63      9.39
06:05:01 AM       em4      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
06:05:01 AM        lo     34.72     34.72      6.95      6.95      0.00      0.00      0.00      0.00
06:05:01 AM       em1   6354.87   7485.17   6526.32  10797.33      0.00      0.00      1.13      8.85
06:05:01 AM       em3      0.00      0.00      0.00      0.00      0.00      0.00      0.00      0.00
 
Cc: davidri...@chromium.org johndhong@chromium.org
Owner: haoweiw@chromium.org
Haowei is this expected?
Summary: chromeos2-devserver5 has 4 ethernet NICs but only 2 have traffic (was: chromes2-devserver5 has 4 ethernet NICs but only 2 have traffic)

Comment 3 by haoweiw@google.com, Feb 6 2017

Yes, this is expected, currently we are not fully utilize all NIC ports on this kind devservers, DELL R710. Only two ports used for bonding NIC. Do we have any concern regarding this configuration? 
No particular concerns.  In the past we found a bug that prevented proper bonding, so I wanted to make sure that this wasn't a similar problem.

It may be helpful to keep those interfaces down if they are not used.  But I don't know if this will cause problems in the future because we're changing some default behavior.

I'll let you decide and close.  Thanks!
Cc: -johndhong@chromium.org englab-sys-cros@google.com
As a FYI...the server is not currently cabled to utilize all 4 ports.  So not sure what changes on your end can interfere with this???

I would prefer that all 4 ports be utilized (although I don't believe we're saturating anyway).  I believe this happened since we had a devserver network cabling convention which was not utilized (or deprecated).

semenzato what was the bug that prevented proper bonding?  I can see a misconfiguration on the server side being possible.
#5: I am having trouble finding the bug.  I looked in crbug, crosbug, and b.

I have emails that refers to the issue from June and July 2016.  The machine was chromeos2-devserver2, and the problem was that the bond0 interface was configure to pick which physical interface to use based on hashing some IP address, rather than hashing on both IP address and MAC address, which is how we fixed the problem.  (Some handwaving here---I don't remember the exact details.)

My earlier comment was referring to the fact that a casual inspection of this devserver may lead one to suspect that bond0 is misconfigured.  But if the unused controller are down, then the casual inspection cannot create this suspicion.

My worry (not sure if founded or unfounded) is that if we configure those interfaces to stay down at boot, this may create confusion in the future when we start using them and don't seem to work as expected.

You would know better.  Thanks.

Comment 7 by aut...@google.com, Feb 7 2017

Labels: -current-issue
Labels: Hotlist-CrOS-DevServerLoad

Comment 9 by haoweiw@google.com, Feb 16 2017

I bounced up the 2 interfaces and now all 4 ports are bonded and active now. 
Status: Fixed (was: Untriaged)
Well that certainly fixes the problem.  Thank you :)

Comment 11 by dchan@google.com, Apr 17 2017

Labels: VerifyIn-59

Comment 12 by dchan@google.com, May 30 2017

Labels: VerifyIn-60
Labels: VerifyIn-61

Comment 14 by dchan@chromium.org, Oct 14 2017

Status: Archived (was: Fixed)

Sign in to add a comment