node['dmi']['system']['manufacturer'] missing #1414

zmscwx · 2019-12-20T15:41:34Z

Description

Software that depends on keys placed directly in node['dmi']['system'] (as opposed to node['dmi']['system']['all_records'][0]) is failing. This affects Ohai (the virtualization plugin, specifically), and by proxy chef-sugar.

Chef Version

15.5.x and later

Platform Version

EL6 and 7

Replication Case

Run chef-shell and check the contents of node['dmi']['system'] - node['dmi']['system']['manufacturer'] should be present.

Client Output

Expected:

# chef-shell
...
chef (15.4.45)> node['dmi']['system']
 => {"all_records"=>[{"record_id"=>"0x0001", "size"=>"1", "application_identifier"=>"System Information", "Manufacturer"=>"VMware, Inc.", "Product Name"=>"VMware Virtual Platform", "Version"=>"None", "Serial Number"=>"VMware-42 2e 6b 5a df 01 5f a7-79 60 2f 6c 6c 40 cd 73", "UUID"=>"5a6b2e42-01df-a75f-7960-2f6c6c40cd73", "Wake-up Type"=>"Power Switch", "SKU Number"=>"Not Specified", "Family"=>"Not Specified"}], "manufacturer"=>"VMware, Inc.", "product_name"=>"VMware Virtual Platform", "version"=>"None", "serial_number"=>"VMware-42 2e 6b 5a df 01 5f a7-79 60 2f 6c 6c 40 cd 73", "uuid"=>"5a6b2e42-01df-a75f-7960-2f6c6c40cd73", "wake_up_type"=>"Power Switch", "sku_number"=>"Not Specified", "family"=>"Not Specified"}

Actual:

# chef-shell
...
chef (15.6.10)> node['dmi']['system']
 => {"all_records"=>[{"record_id"=>"0x0001", "size"=>"1", "application_identifier"=>"System Information", "Manufacturer"=>"VMware, Inc.", "Product Name"=>"VMware Virtual Platform", "Version"=>"None", "Serial Number"=>"VMware-42 2e 2a a1 1b 43 c1 1c-11 30 7f d3 01 b4 aa 3f", "UUID"=>"a12a2e42-431b-1cc1-1130-7fd301b4aa3f", "Wake-up Type"=>"Power Switch", "SKU Number"=>"Not Specified", "Family"=>"Not Specified"}]}

Stacktrace

The text was updated successfully, but these errors were encountered:

tas50 · 2019-12-20T17:57:12Z

@zmscwx Can you post the output of dmidecode on this host?

zmscwx · 2019-12-20T18:04:46Z

I don't think the /full/ output would be feasible, since it's more than I can fit in my scrollback buffer. The pertinent bits are here:

"broken" system

# dmidecode | grep -i vmware
	Manufacturer: VMware, Inc.
	Product Name: VMware Virtual Platform
	Serial Number: VMware-42 XX XX XX XX XX XX XX-XX XX XX XX XX XX XX 3f
	Description: VMware SVGA II

"good" system

# dmidecode | grep -i vmware
	Manufacturer: VMware, Inc.
	Product Name: VMware Virtual Platform
	Serial Number: VMware-42 XX XX XX XX XX XX XX-XX XX XX XX XX XX XX 73
	Description: VMware SVGA II

sshock · 2019-12-29T01:02:09Z

@zmscwx I think it would be helpful if you could paste the entire output of dmidecode, since the most likely reason for those to be missing is if dmidecode were returning multiple results for System Information (but that didn't match).

Alternatively, run ohai directly to get the dmi system section of the json. E.g., on my system I did this and it is working as expected:

git clone [email protected]:chef/ohai.git
cd ohai
bundle
rvmsudo bin/ohai dmi/system

{"all_records"=>
  [{"record_id"=>"0x0001",
    "size"=>"1",
    "application_identifier"=>"System Information",
    "Manufacturer"=>"Shuttle Inc.",
    "Product Name"=>"DS61",
    "Version"=>"V1.0",
    "Serial Number"=>"To be filled by O.E.M.",
    "UUID"=>"03000200-0400-0500-0006-000700080009",
    "Wake-up Type"=>"Power Switch",
    "SKU Number"=>"1.0",
    "Family"=>"D"}],
 "manufacturer"=>"Shuttle Inc.",
 "product_name"=>"DS61",
 "version"=>"V1.0",
 "serial_number"=>"To be filled by O.E.M.",
 "uuid"=>"03000200-0400-0500-0006-000700080009",
 "wake_up_type"=>"Power Switch",
 "sku_number"=>"1.0",
 "family"=>"D"}

zmscwx · 2019-12-30T16:40:20Z

@sshock I can confirm that /ohai/ works as intended - it's specifically in the context of the Chef run that dmi.system appears to be getting clobbered. You can see even in my simple repro case that I had to resort to using chef-shell.

To be perfectly clear, ohai dmi/system returned (my test instances got clobbered over the holidays...) the expected output on both instances. That leads me to believe that this issue doesn't lie within ohai, rather it's something to do with attribute filtering inside of chef/chef (?).

If you still think the full output of dmidecode would be useful, I'll be glad to upload it. Let me know, and I'll spin up some more test instances.

sshock · 2019-12-30T20:41:13Z

@zmscwx Oh, I didn't realize it is working fine for you with just ohai. In that case I don't think we need to see the full output of dmidecode, but at least the (working) output from ohai dmi/system might be helpful.

@tas50 Do you want to transfer this back to chef/chef

zmscwx · 2020-01-13T18:15:35Z

Sorry - I must have missed that last response. Here's the chef-shell output - Note that when compared with the ohai output below, "node['dmi']['system']['manufacturer']" is missing.

chef (15.6.19)> pp node['dmi']['system']
{"all_records"=>
  [{"record_id"=>"0x0001",
    "size"=>"1",
    "application_identifier"=>"System Information",
    "Manufacturer"=>"VMware, Inc.",
    "Product Name"=>"VMware Virtual Platform",
    "Version"=>"None",
    "Serial Number"=>"VMware-42 2e 37 e6 f9 58 36 58-0c f2 6b 30 d4 0d 19 6e",
    "UUID"=>"e6372e42-58f9-5836-0cf2-6b30d40d196e",
    "Wake-up Type"=>"Power Switch",
    "SKU Number"=>"Not Specified",
    "Family"=>"Not Specified"}]}

vs.

$ sudo ohai dmi/system
{
  "all_records": [
    {
      "record_id": "0x0001",
      "size": "1",
      "application_identifier": "System Information",
      "Manufacturer": "VMware, Inc.",
      "Product Name": "VMware Virtual Platform",
      "Version": "None",
      "Serial Number": "VMware-42 2e 37 e6 f9 58 36 58-0c f2 6b 30 d4 0d 19 6e",
      "UUID": "e6372e42-58f9-5836-0cf2-6b30d40d196e",
      "Wake-up Type": "Power Switch",
      "SKU Number": "Not Specified",
      "Family": "Not Specified"
    }
  ],
  "manufacturer": "VMware, Inc.",
  "product_name": "VMware Virtual Platform",
  "version": "None",
  "serial_number": "VMware-42 2e 37 e6 f9 58 36 58-0c f2 6b 30 d4 0d 19 6e",
  "uuid": "e6372e42-58f9-5836-0cf2-6b30d40d196e",
  "wake_up_type": "Power Switch",
  "sku_number": "Not Specified",
  "family": "Not Specified"
}

Sliim · 2020-01-16T12:30:36Z

I have the same issue and I think this is caused by :
https://github.com/chef/ohai/blob/v15.7.3/lib/ohai/common/dmi.rb#L120

Added a debug log to print records.class.to_s and we get:

In ohai:
- records.class.to_s == Mash
In chef-client:
- records.class.to_s == ChefUtils::Mash

The dmi plugin retrieve same informations with dmidecode from ohai and chef-client but the Ohai::Common::DMI.convenience_keys method don't process records due to invalid class name in chef-client run.

zmscwx · 2020-01-16T12:48:23Z

Interesting. Would updating chef/ohai to use records.class.kind_of? Mash be the more correct approach?

Sliim · 2020-01-16T12:50:29Z

This patch looks to fix the issue:

diff --git a/lib/ohai/common/dmi.rb b/lib/ohai/common/dmi.rb
index c7f8a5e1..12c166c5 100644
--- a/lib/ohai/common/dmi.rb
+++ b/lib/ohai/common/dmi.rb
@@ -116,13 +116,14 @@ module Ohai
       # for multiple occurrences of same type, copy to top level all fields and values that are common to all records
       def convenience_keys(dmi)
         dmi.each do |type, records|
+          Ohai::Log.debug("OHAI DEBUG: RECORDS/CLASS: #{records.class.to_s}")
           in_common = Mash.new
-          next unless records.class.to_s == "Mash"
+          next unless ['Mash', 'ChefUtils::Mash'].include?(records.class.to_s)
           next unless records.key?("all_records")
 
           records[:all_records].each do |record|
             record.each do |field, value|
-              next if value.class.to_s == "Mash"
+              next if ['Mash', 'ChefUtils::Mash'].include?(value.class.to_s)
               next if field.to_s == "application_identifier"
               next if field.to_s == "size"
               next if field.to_s == "record_id"

Sliim · 2020-01-16T12:59:52Z

@zmscwx your approach looks better, this also fix the issue:

diff --git a/lib/ohai/common/dmi.rb b/lib/ohai/common/dmi.rb
index c7f8a5e1..140af937 100644
--- a/lib/ohai/common/dmi.rb
+++ b/lib/ohai/common/dmi.rb
@@ -116,13 +116,14 @@ module Ohai
       # for multiple occurrences of same type, copy to top level all fields and values that are common to all records
       def convenience_keys(dmi)
         dmi.each do |type, records|
+          Ohai::Log.debug("OHAI DEBUG: RECORDS/CLASS: #{records.class.to_s}")
           in_common = Mash.new
-          next unless records.class.to_s == "Mash"
+          next unless records.kind_of? Mash
           next unless records.key?("all_records")
 
           records[:all_records].each do |record|
             record.each do |field, value|
-              next if value.class.to_s == "Mash"
+              next if value.kind_of? Mash
               next if field.to_s == "application_identifier"
               next if field.to_s == "size"
               next if field.to_s == "record_id"

zmscwx · 2020-01-20T14:07:45Z

@sshock @tas50 do either of you have input on this approach (using kind_of? in chef/ohai)? My gut says that since many other projects depend on Ohai, changes should be approached with caution.

tas50 · 2020-01-20T17:51:57Z

We really should have been using kind_of (or is_a) the whole time. This approach seems entirely appropriate to me. Thanks for digging into this.

#1423

zmscwx · 2020-01-20T17:53:10Z

gladly! thank you for the quick turnaround.

Sliim · 2020-01-20T18:17:31Z

👍

tas50 transferred this issue from chef/chef Dec 20, 2019

zmscwx closed this as completed Jan 20, 2020

comay mentioned this issue Mar 12, 2023

Another instance of possible incorrect class comparison #1790

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

node['dmi']['system']['manufacturer'] missing #1414

node['dmi']['system']['manufacturer'] missing #1414

zmscwx commented Dec 20, 2019

tas50 commented Dec 20, 2019

zmscwx commented Dec 20, 2019

sshock commented Dec 29, 2019 •

edited

Loading

zmscwx commented Dec 30, 2019

sshock commented Dec 30, 2019

zmscwx commented Jan 13, 2020

Sliim commented Jan 16, 2020

zmscwx commented Jan 16, 2020

Sliim commented Jan 16, 2020

Sliim commented Jan 16, 2020

zmscwx commented Jan 20, 2020

tas50 commented Jan 20, 2020

zmscwx commented Jan 20, 2020

Sliim commented Jan 20, 2020

node['dmi']['system']['manufacturer'] missing #1414

node['dmi']['system']['manufacturer'] missing #1414

Comments

zmscwx commented Dec 20, 2019

Description

Chef Version

Platform Version

Replication Case

Client Output

Stacktrace

tas50 commented Dec 20, 2019

zmscwx commented Dec 20, 2019

sshock commented Dec 29, 2019 • edited Loading

zmscwx commented Dec 30, 2019

sshock commented Dec 30, 2019

zmscwx commented Jan 13, 2020

Sliim commented Jan 16, 2020

zmscwx commented Jan 16, 2020

Sliim commented Jan 16, 2020

Sliim commented Jan 16, 2020

zmscwx commented Jan 20, 2020

tas50 commented Jan 20, 2020

zmscwx commented Jan 20, 2020

Sliim commented Jan 20, 2020

sshock commented Dec 29, 2019 •

edited

Loading