Bug 1347548
| Summary: | It can't start rdma service during RHEL-7.x PXE installation over IPoIB | ||||||
|---|---|---|---|---|---|---|---|
| Product: | Red Hat Enterprise Linux 7 | Reporter: | zguo <zguo> | ||||
| Component: | rdma | Assignee: | Jarod Wilson <jarod> | ||||
| Status: | CLOSED ERRATA | QA Contact: | zguo <zguo> | ||||
| Severity: | high | Docs Contact: | |||||
| Priority: | high | ||||||
| Version: | 7.3 | CC: | abeausol, ddutile, dledford, honli, jshortt, noas, trinh.dao, zguo | ||||
| Target Milestone: | rc | ||||||
| Target Release: | --- | ||||||
| Hardware: | x86_64 | ||||||
| OS: | Linux | ||||||
| Whiteboard: | |||||||
| Fixed In Version: | Doc Type: | If docs needed, set a value | |||||
| Doc Text: | Story Points: | --- | |||||
| Clone Of: | Environment: | ||||||
| Last Closed: | 2016-11-04 03:26:30 UTC | Type: | Bug | ||||
| Regression: | --- | Mount Type: | --- | ||||
| Documentation: | --- | CRM: | |||||
| Verified Versions: | Category: | --- | |||||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||
| Cloudforms Team: | --- | Target Upstream Version: | |||||
| Embargoed: | |||||||
| Bug Depends On: | 1348381 | ||||||
| Bug Blocks: | 1213541, 1274397 | ||||||
| Attachments: |
|
||||||
|
Description
zguo
2016-06-17 08:33:06 UTC
(In reply to zguo from comment #0) ... > While installing RHEL-7.3-20160615.n.0 via PXE over IPoIB > Message from console output: > [FAILED] Failed to start Initialize the iWAR...niBand/RDMA stack in the > kernel. Please try generating an install image using this updated rdma test package: http://hp-dl360pgen8-07.khw.lab.eng.bos.redhat.com/~jwilson/misc/rdma/rdma-7.3_4.7_rc2-3.el7.noarch.rpm Basic gist of the change in the updated test package linked in the prior comment:
diff --git a/rdma.modules-setup.sh b/rdma.modules-setup.sh
index ca9322d..15f1627 100644
--- a/rdma.modules-setup.sh
+++ b/rdma.modules-setup.sh
@@ -26,4 +26,5 @@ install() {
installkernel() {
hostonly='' instmods =drivers/infiniband =drivers/net/ethernet/mellanox =drivers/net/ethernet/chelsio =drivers/net/ethernet/cisco =drive
+ hostonly='' instmods target_core_mod crc-t10dif crct10dif_common
}
(In reply to Jarod Wilson from comment #4) > Basic gist of the change in the updated test package linked in the prior > comment: > At least, it is not a completed solution for this issue. The kernel/target drivers will be removed via /usr/share/lorax/runtime-cleanup.tmpl. # vim /usr/share/lorax/runtime-cleanup.tmpl 83 ## clean up kernel modules 84 removekmod sound drivers/media drivers/hwmon \ 85 net/atm net/bluetooth net/sched net/sctp \ 86 net/rds net/l2tp net/decnet net/netfilter net/ipv4 net/ipv6 \ 87 drivers/watchdog drivers/target drivers/rtc drivers/input/joystick \ 88 drivers/bluetooth drivers/hid drivers/edac \ 89 drivers/usb/serial drivers/usb/host drivers/usb/misc \ 90 fs/ocfs2 fs/ceph fs/nfsd fs/ubifs fs/nilfs2 \ 91 arch/x86/kvm Managed to start rdma service.
[root@ib2-qa-03 ~]# diff -Nurp /usr/share/lorax/runtime-cleanup.tmpl.bak /usr/share/lorax/runtime-cleanup.tmpl
--- /usr/share/lorax/runtime-cleanup.tmpl.bak 2016-06-19 10:07:08.540835900 -0400
+++ /usr/share/lorax/runtime-cleanup.tmpl 2016-06-19 10:07:36.257351230 -0400
@@ -84,7 +84,7 @@ remove /usr/share/icons/*/icon-theme.cac
removekmod sound drivers/media drivers/hwmon \
net/atm net/bluetooth net/sched net/sctp \
net/rds net/l2tp net/decnet net/netfilter net/ipv4 net/ipv6 \
- drivers/watchdog drivers/target drivers/rtc drivers/input/joystick \
+ drivers/watchdog drivers/rtc drivers/input/joystick \
drivers/bluetooth drivers/hid drivers/edac \
drivers/usb/serial drivers/usb/host drivers/usb/misc \
fs/ocfs2 fs/ceph fs/nfsd fs/ubifs fs/nilfs2 \
[root@ib2-qa-03 ~]#
[anaconda root@localhost ~]# cat /lib/dracut/modules.d/05rdma/module-setup.sh
#!/bin/bash
check() {
[ -n "$hostonly" -a -d /sys/class/infiniband_verbs/uverbs0 ] && return 0
[ -n "$hostonly" ] && return 255
return 0
}
depends() {
return 0
}
install() {
inst /etc/rdma/rdma.conf
inst /etc/rdma/mlx4.conf
inst /etc/rdma/sriov-vfs
inst /usr/libexec/rdma-init-kernel
inst /usr/libexec/rdma-fixup-mtrr.awk
inst /usr/libexec/mlx4-setup.sh
inst /usr/libexec/rdma-set-sriov-vf
inst /usr/lib/modprobe.d/libmlx4.conf
inst_multiple lspci setpci awk sleep
inst_multiple -o /etc/modprobe.d/mlx4.conf
inst_rules 98-rdma.rules 70-persistent-ipoib.rules
}
installkernel() {
hostonly='' instmods =drivers/infiniband =drivers/net/ethernet/mellanox =drivers/net/ethernet/chelsio =drivers/net/ethernet/cisco =drivers/net/ethernet/emulex =drivers/target
hostonly='' instmods crc-t10dif crct10dif_common
}
[anaconda root@localhost ~]#
[anaconda root@localhost ~]# systemctl status rdma.service -l
● rdma.service - Initialize the iWARP/InfiniBand/RDMA stack in the kernel
Loaded: loaded (/usr/lib/systemd/system/rdma.service; disabled; vendor preset: disabled)
Active: active (exited) since Sun 2016-06-19 14:32:37 UTC; 1min 35s ago
Docs: file:/etc/rdma/rdma.conf
Process: 2046 ExecStart=/usr/libexec/rdma-init-kernel (code=exited, status=0/SUCCESS)
Main PID: 2046 (code=exited, status=0/SUCCESS)
CGroup: /system.slice/rdma.service
Jun 19 14:32:36 localhost systemd[1]: Starting Initialize the iWARP/InfiniBand/RDMA stack in the kernel...
Jun 19 14:32:37 localhost systemd[1]: Started Initialize the iWARP/InfiniBand/RDMA stack in the kernel.
(In reply to Honggang LI from comment #7) > Managed to start rdma service. > > [root@ib2-qa-03 ~]# diff -Nurp /usr/share/lorax/runtime-cleanup.tmpl.bak > /usr/share/lorax/runtime-cleanup.tmpl > --- /usr/share/lorax/runtime-cleanup.tmpl.bak 2016-06-19 10:07:08.540835900 > -0400 > +++ /usr/share/lorax/runtime-cleanup.tmpl 2016-06-19 10:07:36.257351230 -0400 > @@ -84,7 +84,7 @@ remove /usr/share/icons/*/icon-theme.cac > removekmod sound drivers/media drivers/hwmon \ > net/atm net/bluetooth net/sched net/sctp \ > net/rds net/l2tp net/decnet net/netfilter net/ipv4 net/ipv6 \ > - drivers/watchdog drivers/target drivers/rtc > drivers/input/joystick \ > + drivers/watchdog drivers/rtc drivers/input/joystick \ > drivers/bluetooth drivers/hid drivers/edac \ > drivers/usb/serial drivers/usb/host drivers/usb/misc \ > fs/ocfs2 fs/ceph fs/nfsd fs/ubifs fs/nilfs2 \ Okay, so we need another bug for another component here. > [anaconda root@localhost ~]# cat > /lib/dracut/modules.d/05rdma/module-setup.sh > #!/bin/bash > > check() { > [ -n "$hostonly" -a -d /sys/class/infiniband_verbs/uverbs0 ] && return 0 ... > installkernel() { > hostonly='' instmods =drivers/infiniband =drivers/net/ethernet/mellanox > =drivers/net/ethernet/chelsio =drivers/net/ethernet/cisco > =drivers/net/ethernet/emulex =drivers/target > hostonly='' instmods crc-t10dif crct10dif_common This wasn't from the test package I requested to be tested. The -d in check() is wrong (it should be -c), and I included only what I believe was the only target module needed, target_core_mod. But good that we're getting close. Is anything else under drivers/target/ actually required here? (In reply to Jarod Wilson from comment #8) > (In reply to Honggang LI from comment #7) > Okay, so we need another bug for another component here. zguo, please file a lorax bug and set it as a blocker of this bug. > > This wasn't from the test package I requested to be tested. No, it is not your test package. I tried it. Failed to load ib_srpt module because of missing target modules. And lorax runtime-cleanup.tmpl file removes entire drivers/target directory. So, I dropped your and overwrite the old rdma package with my hacked one. > The -d in > check() is wrong (it should be -c), Yes, you are right. That is why I force lorax to install rdma with 'lorax -i rdma blabla ....'. > and I included only what I believe was > the only target module needed, target_core_mod. But good that we're getting > close. Is anything else under drivers/target/ actually required here? I'd suggest to package up drivers/target. When pxeboot/initrd.img and LiveOS/squashfs.img have different modules, it is really confusing for debugging. (In reply to Honggang LI from comment #9) > (In reply to Jarod Wilson from comment #8) > > (In reply to Honggang LI from comment #7) > > > Okay, so we need another bug for another component here. > > zguo, please file a lorax bug and set it as a blocker of this bug. https://bugzilla.redhat.com/show_bug.cgi?id=1348381 > > > > This wasn't from the test package I requested to be tested. > > No, it is not your test package. I tried it. Failed to load ib_srpt module > because of missing target modules. And lorax runtime-cleanup.tmpl file > removes entire drivers/target directory. So, I dropped your and overwrite > the old rdma package with my hacked one. > > > The -d in > > check() is wrong (it should be -c), > > Yes, you are right. That is why I force lorax to install rdma with 'lorax -i > rdma blabla ....'. > > > and I included only what I believe was > > the only target module needed, target_core_mod. But good that we're getting > > close. Is anything else under drivers/target/ actually required here? > > I'd suggest to package up drivers/target. When pxeboot/initrd.img and > LiveOS/squashfs.img have different modules, it is really confusing for > debugging. (In reply to Honggang LI from comment #9) > (In reply to Jarod Wilson from comment #8) ... > > This wasn't from the test package I requested to be tested. > > No, it is not your test package. I tried it. Failed to load ib_srpt module > because of missing target modules. And lorax runtime-cleanup.tmpl file > removes entire drivers/target directory. So, I dropped your and overwrite > the old rdma package with my hacked one. That was the feedback I was looking for when asking to test the package I created. :) I'm not quite sure why that change didn't work, based on: $ modinfo ib_srpt |grep target description: InfiniBand SCSI RDMA Protocol target v2.0.0 (2011-02-14) depends: ib_core,target_core_mod,ib_cm,ib_mad All of those should have been present and satisfied. > > and I included only what I believe was > > the only target module needed, target_core_mod. But good that we're getting > > close. Is anything else under drivers/target/ actually required here? > > I'd suggest to package up drivers/target. When pxeboot/initrd.img and > LiveOS/squashfs.img have different modules, it is really confusing for > debugging. Despite the head-scratching about ib_srpt failing, since iser looks to need more of drivers/target/* anyway, and it's working for you, I'll add all of drivers/target/*. (In reply to Jarod Wilson from comment #11) > Despite the head-scratching about ib_srpt failing, since iser looks to need > more of drivers/target/* anyway, and it's working for you, I'll add all of > drivers/target/*. Thank you! Issue is gone with latest rdma package.
Check the rdma service after installation:
[root@localhost log]# service rdma status
Redirecting to /bin/systemctl status rdma.service
● rdma.service - Initialize the iWARP/InfiniBand/RDMA stack in the kernel
Loaded: loaded (/usr/lib/systemd/system/rdma.service; disabled; vendor preset: disabled)
Active: active (exited) since Wed 2016-07-27 05:24:15 EDT; 3min 32s ago
Docs: file:/etc/rdma/rdma.conf
Process: 699 ExecStart=/usr/libexec/rdma-init-kernel (code=exited, status=0/SUCCESS)
Main PID: 699 (code=exited, status=0/SUCCESS)
CGroup: /system.slice/rdma.service
Jul 27 05:24:11 localhost.localdomain systemd[1]: Starting Initialize the iWA...
Jul 27 05:24:15 localhost.localdomain systemd[1]: Started Initialize the iWAR...
Hint: Some lines were ellipsized, use -l to show in full.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://rhn.redhat.com/errata/RHEA-2016-2309.html Hi Andre, Any plans to have this fix for RHEL 7.2 as well? What is the status for RHEL 7.1? Thanks, Noa The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days |