Hi Lee, Thanks for looking into this. Yes, this happens under excess load. It's hit rarely (~1 in 100) and does look like loss of some event in specific scenario. Would update back if we get more info.
Thanks, Satyajit On Monday, January 21, 2019 at 10:33:52 AM UTC-8, The Lee-Man wrote: > > On Monday, December 17, 2018 at 9:08:54 AM UTC-8, Satyajit Deshmukh wrote: >> >> >> >> On Sunday, December 16, 2018 at 5:30:29 PM UTC-8, The Lee-Man wrote: >>> >>> On Friday, December 14, 2018 at 12:13:46 PM UTC-8, Satyajit Deshmukh >>> wrote: >>>> >>>> Hello, >>>> >>>> An update on the issue. I could observe that the target entries were >>>> not populated under sysfs. >>>> >>>> This is for a session that has a valid block device: >>>> $ ls /sys/class/iscsi_session/session778/device >>>> connection778:0 iscsi_session power target162:0:0 uevent >>>> >>>> > I am trying to reproduce your errors myself, but so far no success. > > It sounds like what is occurring is that some event or sequence of steps > is failing under your conditions. So I need to reproduce your conditions as > closely as possible. > > I've created 75 targets on my target host system and connected to them > repeatedly, but so far none of the disc devices has failed to show up. I > suspect it's timing- and/or load-related. And how often are you seeing > these "no disc device" events relative to normal behavior? > >> This is for a session that does not have a valid block device: >>>> $ls /sys/class/iscsi_session/session780/device >>>> connection780:0 iscsi_session power uevent >>>> >>>> As we can see, the target... directory is missing. >>>> So, an event responsible to create the sysfs entry could not get >>>> created. >>>> >>>> journalctl does not print this info. Is there a way to enable some >>>> debugging, to debug this? >>>> >>>> >>>> >>> It seems like the iscsi initiator code in the kernel is not creating the >>> target directory. I will have to look at the code to figure out why. Is >>> there any difference between the two targets? How many targets to you have? >>> What type of targets are they (i.e. hardware, software)? >>> >>> >> There is no difference between the two targets. We have 100s of iSCSI >> targets on a single VM. All of these are software targets. >> The target device does get created most of the times. >> >> Another related issue we found is during log outs. In that scenario, the >> block device was not cleanly removed, during the iscsiadm logout command. I >> will share details about that shortly. >> > -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/open-iscsi. For more options, visit https://groups.google.com/d/optout.
