topic Re: RHEL 3.0 Update 8 kernel (?) problems in network (?) stack in Operating System - Linux

RHEL 3.0 Update 8 kernel (?) problems in network (?) stack

TAck_1 — Wed, 13 Sep 2006 23:24:04 GMT

We recently began testing RHEL 3.0 Update 8 (which includes a kernel update from 2.4.21-40 (update 7) to 2.4.21-47) and are encountering sporadic problems across both test systems we upgraded.

Both systems run Oracle databases and versions of Oracle application server (10AS v1 and 9iAS v1).

The symptoms we see seem to center on network stack problems. We see Oracle forms server processes hanging (apparently on network calls according to strace data). We see the inability to start Oracle forms builder (strace data indicates failed socket connections). We see outbound SMTP transfers failing. We sometimes see hangs when doing df's. We sometimes see failures as the upgraded system attempts to access NFS volumes from other RHEL 3.0 systems.

We can always clear the problems by rebooting the system - the system will run fine for 1 to 2 days, then enter failure mode again. We can avoid the problems by booting to the old kernel (Update 7 - 2.4.21-40) without backing out the other RPM's from update 8.

We are working cases with HP support and Oracle support, but since we can't recreate the conditions at will and because the symptoms change slightly with each occurrence, progress is slow.

I have searched for other reports of such problems with the 2.4.21-47 kernel, but haven't found any. I am wondering if the problem may be unique to kernel modules we have loaded - we run Proliant DL580 G2 systems with QLogic FC host bus adapters and HP SIM agents. So our kernel has modules for the HP specific stuff. We insured the HP kernel modules were being used and loaded (for things like the QLogic driver, which RedHat also delivers).

Has anyone else see any such problems with the 2.4.21-47 kernel or RHEL 3.0 Update 8?
Are folks successfully running Proliant hardware with QLogic FC HBA's and SIM agents on the 2.4.21-47 kernel and RHEL 3.0 Update 8?

Thanks.

Re: RHEL 3.0 Update 8 kernel (?) problems in network (?) stack

Ragu_3 — Sun, 24 Sep 2006 23:13:45 GMT

Does this have the Broadcom Gigabit NIC? This ethernet card has known issues with the specific Redhat v2.4 Linux kernel, network stack issues. Can you get the newer v2.6.18 kernel and re-compile on your production machine? Look out and do this only if you are sure. Then you should select the tg3 (Tigon card) network card driver during 'make menuconfig'.

Re: RHEL 3.0 Update 8 kernel (?) problems in network (?) stack

George Liu_4 — Mon, 25 Sep 2006 11:55:55 GMT

We don't have any issues on RHEL 3 Update 8 on Proliant DL145/380/385/.
If you are certain the old kernel works but 2.4.21-47 breaks, I don't see any reason you couldn't use the old kernel.

Re: RHEL 3.0 Update 8 kernel (?) problems in network (?) stack

TAck_1 — Mon, 25 Sep 2006 13:21:59 GMT

The NICs are Intel 82546EB GigE controllers, so not Broadcom. Our path of last resort would be to go to a 2.6 kernel (which would translate into a RHEL 4.0 upgrade which we would do via a reinstall).

Our primary driver to go to the latest kernel (RHEL 3 Update 8) is to address kernel security vulnerabilities, so while the older kernel works, we need to move forward to keep current with security patches.

Thanks.