Approximate one minute downtime is expected in case of logical network failure which provides the network to virtual server. (See [Service Descriptions - Support] for detail.)
If you set up storage connection by using iSCSI from virtual server, please take account into the downtime.
If you connect to our block storage service, please refer [Tutorials - Attaching resources to a Block Storage Volume].
If you connect to the storage that we don't provide, please contact your storage provider.
The following shows some examples of configuration to hold I/O during the downtime.
The impact on your application has to be tested and evaluated additionally for your system.
The multipath devices can be used to make I/O queued at multipath layer in case of disconnection to iSCSI target.
For RHEL guest, the option of 'features "1 queue_if_no_path"' makes I/Os held at multi-path layer when all paths are down.
If the multipath devices are not used, the timeout of iSCSI layer needs to be set long enough.
For RHEL guest, we set iSCSI initiator timeout to 120 seconds.
Configuration file name: /etc/iscsi/iscsid.conf
Parameter: node.session.timeo.replacement_timeout
Value: 120 (Default value: 120)
For Windows guest, we set both disks I/O timeout and iSCSI initiator request timeout to 120 seconds.
Key: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\disk
Name: TimeoutValue
Value: 120 (Default value: 60)
Key: HKLM\SYSTEM\CurrentControlSet\Control\Class{4D36E97B-E325-11CE-BFC1-08002BE10318}\0000\Parameters (*1)
Name: MaxRequestHoldTime
Value: 120 (Default value: 60)
(*1) The part of "0000" may vary. Please choose the instance number of Microsoft iSCSI Initiator.