Sun Dec 24, 2017 8:18 pm
Serviio w/ ESXi and iSCSI Round Robin issues
I've recently migrated my Serviio setup to a Windows 2016 server running on ESXi 6.5.
The Windows server has it's root drive (local host machine drive) and then a mounted datastore on a separate drive. That drive is a datastore which is located on a dell PS6000e SAN.
The Dell SAN has 3 of the 4 network connections dedicated to iSCSI traffic only (the 4th is the management interface) which is connected to a Cisco SG200-26 1GB switch.
Jumbo frames are enabled on the SAN, switch and in VMware
Traffic on iSCSI ports are on a separate iSCSI VLAN
Anytime I go into the ESXi configuration and change the path selection on an iSCSI target from Fixed (VMware) to Round Robin (VMware) Serviio will play a media file for about 10-15 seconds and then the device gets disconnected and Serviio is no longer able to be discovered on the network until I restart the Windows server (after the previous occurs twice in a row). As soon as I change the settings back to fixed so that only a single connection is being used, Serviio gives me no issues and it can run for ages w/o issue.
I doubt it's a performance issue as the server has been assigned 4 CPU's (single core per socket) with 8GB of memory and I've not seen the performance monitor max out.
The round robin configuration works with everything else that reads and writes to the SAN (I've got graphs of copying 400GB of data to it showing the transfer of data balanced across the different network connections).
I've enabled the debug logs and re-run the crash scenario. The definitive line seems to be this:
2017-12-24 14:06:52,777 DEBUG [WebServer] I/O error: Connection reset by peer: socket write error
Anyone seen anything similar, have suggestions etc...?
If need be, I can always zip up the logs.
[EDIT]meant to mention that to ensure it wasn't a network issue, I simultaneously tested copying files to the same datastore across the network links and there was no sign of issue there and the status for all paths in VMware remained in Active state.
The Windows server has it's root drive (local host machine drive) and then a mounted datastore on a separate drive. That drive is a datastore which is located on a dell PS6000e SAN.
The Dell SAN has 3 of the 4 network connections dedicated to iSCSI traffic only (the 4th is the management interface) which is connected to a Cisco SG200-26 1GB switch.
Jumbo frames are enabled on the SAN, switch and in VMware
Traffic on iSCSI ports are on a separate iSCSI VLAN
Anytime I go into the ESXi configuration and change the path selection on an iSCSI target from Fixed (VMware) to Round Robin (VMware) Serviio will play a media file for about 10-15 seconds and then the device gets disconnected and Serviio is no longer able to be discovered on the network until I restart the Windows server (after the previous occurs twice in a row). As soon as I change the settings back to fixed so that only a single connection is being used, Serviio gives me no issues and it can run for ages w/o issue.
I doubt it's a performance issue as the server has been assigned 4 CPU's (single core per socket) with 8GB of memory and I've not seen the performance monitor max out.
The round robin configuration works with everything else that reads and writes to the SAN (I've got graphs of copying 400GB of data to it showing the transfer of data balanced across the different network connections).
I've enabled the debug logs and re-run the crash scenario. The definitive line seems to be this:
2017-12-24 14:06:52,777 DEBUG [WebServer] I/O error: Connection reset by peer: socket write error
Anyone seen anything similar, have suggestions etc...?
If need be, I can always zip up the logs.
[EDIT]meant to mention that to ensure it wasn't a network issue, I simultaneously tested copying files to the same datastore across the network links and there was no sign of issue there and the status for all paths in VMware remained in Active state.
