Fix fibrechannel_linux for ppc64le

[hamzy]

Seen on a master node:

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x106cefa4]
goroutine 81 [running]:
github.com/prometheus/node_exporter/collector.(*fibrechannelCollector).Update(0x7e28c1ae20f0, 0x7e28c1826770)
  /go/src/github.com/prometheus/node_exporter/collector/fibrechannel_linux.go:133 +0x604

sh-5.1# ls -l /host/sys/class/fc_host/
total 0
lrwxrwxrwx. 1 root root 0 Jul 30 20:42 host1 -> ../../devices/vio/30000003/host1/fc_host/host1
lrwxrwxrwx. 1 root root 0 Jul 30 20:42 host2 -> ../../devices/vio/30000005/host2/fc_host/host2
sh-5.1# ls -l /host/sys/class/fc_host/host1/
total 0
-rw-r--r--. 1 root root 65536 Jul 30 20:09 dev_loss_tmo
lrwxrwxrwx. 1 root root     0 Jul 30 20:42 device -> ../../../host1
-r--r--r--. 1 root root 65536 Jul 30 20:09 fabric_name
--w-------. 1 root root 65536 Jul 30 20:42 issue_lip
-r--r--r--. 1 root root 65536 Jul 30 20:42 maxframe_size
-r--r--r--. 1 root root 65536 Jul 30 20:09 node_name
-r--r--r--. 1 root root 65536 Jul 30 20:09 port_id
-r--r--r--. 1 root root 65536 Jul 30 20:42 port_name
-r--r--r--. 1 root root 65536 Jul 30 20:09 port_state
-r--r--r--. 1 root root 65536 Jul 30 20:09 port_type
drwxr-xr-x. 2 root root     0 Jul 30 20:42 power
-r--r--r--. 1 root root 65536 Jul 30 20:09 speed
lrwxrwxrwx. 1 root root     0 Jul 30 20:42 subsystem -> ../../../../../../class/fc_host
-r--r--r--. 1 root root 65536 Jul 30 20:09 supported_classes
-rw-r--r--. 1 root root 65536 Jul 30 20:42 tgtid_bind_type
-rw-r--r--. 1 root root 65536 Jul 30 20:42 uevent

Since many filenames for counters are not guaranteed to exist in VMs, allocate if not present.

[prb112]

This defense addition makes sense to me.

[prb112]

Related to a bug found at https://redhat.atlassian.net/browse/OCPBUGS-100376

[simonpasquier]

this info metric isn't correct (most of the label values are empty)

[hamzy]

Isn't that the whole point of this PR?

[simonpasquier]

The expected output is

node_fibrechannel_info{dev_loss_tmo="",fabric_name="",fc_host="host1",port_id="",port_name="",port_state="Online",port_type="",speed="8 Gbit",supported_classes="",supported_speeds="",symbolic_name=""} 1                                                                                                                                                                                 
node_fibrechannel_info{dev_loss_tmo="",fabric_name="",fc_host="host2",port_id="",port_name="",port_state="Online",port_type="",speed="",supported_classes="",supported_speeds="",symbolic_name=""} 1                                                                                                                                                                                       
node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host0",port_id="000002",port_name="1000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1                    

instead of

node_fibrechannel_info{dev_loss_tmo="",fabric_name="",fc_host="host1",port_id="",port_name="",port_state="Online",port_type="",speed="8 Gbit",supported_classes="",supported_speeds="",symbolic_name=""} 1
node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host0",port_id="000002",port_name="1000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1
node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host2",port_id="000003",port_name="4000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1

Notice how the host2 line is between host1 and host0 and how most of its labels are empty.

[hamzy]

I still don't understand what the problem is. The testcase reproduces the problem without the fix and everything passes with the fix.

Can you just provide a patch for whatever it is you are unhappy with?

[simonpasquier]

The CI isn't happy, I'll try to get you a patch.

[hamzy]

I appreciate it, thanks!

[prb112]

/lgtm

not on project directly, just reviewing for colleague @hamzy

[hamzy]

@simonpasquier What is the next step in the process?

[jan--f]

The changes in ‎collector/fixtures/sys.ttar look a bit messy and it's hard to distinguish what is really required for this change. Some parts of the diff seems to just reorder content.
Is that needed or can that reordering be dropped?

[hamzy]

This project doesn't seem to have the proper toolset to support just targeted additions. to components. Can you provide some help in this area? I guess what needs to happen is to uncompress the tarball and then git checkout files? But I always feel uncomfortable performing manual changes on a critical component.

[jan--f]

collector/fixtures/sys.ttar is a plain text file. It was initially created with a tool like https://github.com/ideaship/ttar.
So normal git commands work well there. Perhaps just git add -p the parts that are new and drop the reordering.

[jan--f]

Great, thank you!