Description of problem: Kernel hang while running Cyclone (http://www.highwinds- software.com/products/product.aspx?productID=4&pageID=1) System unresponsive. Physical reset required. If the system is configured to handle more than a few feeds and network traffic goes beyond 40mbit/s the system will crash within 10 seconds of starting Cyclone. Fiddled with multiple hardware including: Intel STL2 mobo with dual 1ghz-p3, mobo has intel 100 ethernet Supermicro Superserver 6013P-8 with dual 2,8 Xeon's, w/ & wo/ HT, mobo has dual intel copper gige Storage as 5x IBM 36gb 10k SCSI (IC35L036UCD210-0, rev S5BS) on an external u160 box or in internal slots of Intel SC5000 case. Megaraid Elite 1600 running JBOD or RAID5 Adaptec 7902 w/ & wo/ 2005S ZCR running JBOD or RAID5 Two channel Symbios 896-based u2w-card running JBOD Intel dual port 1000MT gige PCI-X Version-Release number of selected component (if applicable): 2.4.21-9.0.1.ELsmp How reproducible: Every time Steps to Reproduce: 1. Install RHEL3 U1, update to current 2. Get Cyclone, configure more than a few feeds 3. Start Cyclone Actual results: Kernel hang Expected results: Cyclone keeps pumping news and the system keeps running Additional info: Currently running Arjan's 2.6.3-2.1.213smp on STL2 mobo with Symbios, disks in the external case. Connectivity with internal e100. Crashes every few days and spews errors continuously.
Created attachment 98172 [details] Console output
Created attachment 98174 [details] Syslog messages while running 2.6.3-2.1.213smp
Oh yes, tried ext3 with defaults and reiserfs, no difference.
Created attachment 98175 [details] Semi-working 2.6-system crashing with 2.4
I booted the system to 2.4.21-9.0.1.ELsmp. The hardware is now: Intel STL2 mobo with onboard e100. OS-drives connected to the onboard SCSI. Symbios896 handling the news spool.
Well I booted to 2.4.21-9.ELsmp instead it seems. More crash output available if needed, can be generated at will.
I dont really see this is a dpt_i2o problem.
I've been informed that Highwind's assistance with this is required so I'll try pushing them to approach Redhat for certifying their stuff with RHEL.
Works fine on RHEL4.