<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN"> <HTML> <HEAD> <META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=us-ascii"> <META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7226.0"> <TITLE>Problems using multipath-tools & dm-multipath.</TITLE> </HEAD> <BODY>  I am trying to use multipath-tools version 0.2.7 on a system running 2.6.6 kernel with the 2.6.6-rc2-udm1 dm patchset. The system has dual port Qlogic 2312 controller, connected to a JBOD through a FC switch. I have the multipath drives configured as follows - path_grouping_policy multibus - getuid_callout /bin/scsi_id -g -s - path_checker tur I start multipathd; run multipath; run kpartx on a partioned drive dm-7 (this results in additional devmaps dm-11 & dm-12); I then run bonnie++ on the partition dm-11. While bonnie++ is in progress I disconnect one of the FC ports from the switch. I expect -- the path switch to happen and the failed bios to be retried so the application can run after a delay. I find that the application waits for a long while, reports an I/O error and quits. The behaviour is not different with other applications that access the drive. Is this supposed to work ? What could be the problem here ? Secondly I connect the port back to the switch. I expect the path re-connect to be discovered and the multipath device map to be restored. But subsequent disk I/O through the same devmapped device (dm-11) sends data only through one port (one that has always been connected). The path restoration does not happen. Do you have thoughts on these. Appreciate your feedback on this, Ramesh. </BODY> </HTML>