Hi there, MY observations related to this behaviour are:
I could see this message only in virtual machines (VMWare guests) installed with debian/ubuntu. Through the VMWare Forums i could find that there "should" be a gap in the LSI-Drivers from linux. Not sure if its true, i found my own workaround. This warning/error messages from mpt-status comes around with entries in /var/log/kern.log like: kernel: [1659597.085565] mptscsih: ioc0: attempting task abort! (sc=ffff880115ba1000) kernel: [1659597.085575] sd 0:0:0:0: [sda] CDB: Write(10): 2a 00 00 05 58 00 00 00 10 00 kernel: [1659597.085587] mptscsih: ioc0: task abort: SUCCESS (sc=ffff880115ba1000) i figured for ME, that those issues comes only up if the disk load to the storage on the underlying VMWare-System ( VMWareServer VMWareESX, VMWareESXI are tested by me for this case) is a bit higher. The Host containing Linux system moves the latency times for the disk access a bit up and for the emulated LSI-Logic it looks klike the disks have timing issues or are temporary disconnected from the SCSI/SATA-Bus. MY workaround to not run into such trouble: #1 reconfigure the virtual machines with an BusLogic SCSI HBA #2 convert the related virtual disks with vmware-vdiskmanager to BusLogic Disks #3 spread the load over different storages to #3: if you have vm's that access through each other the same disk areas on the same storage and interfering e.g. a FTP-Server delivers per NFS/SMB files to another host laying on the same storage and storing data from those into the same storage area -> gotcha...2 systems fight against each other....i think you got what i mean - concurrent access Another hint ( if you can't spread the data over different storages but if you are able to add more disks): for heavy usage VMWare or any other virtualization system storage or database storage configure your systems with different disks for system and VM-Storage(datastorage) and configure the storages as RAID10 ( be aware not RAID 01!! ) and use as much disks as you can provide with your controllers/budged. if its possible spread the RAID10 ( the RAID1 part) over 2 different SCSI/Disk controllers or SAN/NAS systems. For the RAID10 setup and adding more disks just take a look at the well known documentation or feel free to contact me. just my 20ct. regards m. -- = = = http://michael-schuh.net/ = = =
