auto reconnect

[clicx]

How to set auto reconnect to server if connection broken?
It seems not recovery iscsi connection after host is down for a period time.

[sahlberg]

You can set how many times libiscsi will try to reconnect before giving up using : void iscsi_set_reconnect_max_retries(struct iscsi_context *iscsi, int count) This should however default to -1, which means never give up, try reconnecting forever.

…

On Fri, Jan 5, 2018 at 7:53 PM, clicx ***@***.***> wrote: How to set auto reconnect to server if connection broken? It seems not recovery iscsi connection after host is down for a period time. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#260>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAeNkCnAHSodIqE3VqRH6JWTX6YiquPTks5tHfEngaJpZM4RUPS6> .

[clicx]

iscsi_tcp_service (iscsi=0x643c80, revents=4) at lib/socket.c:868
868 if (iscsi->fd < 0) {
(gdb) p iscsi->fd
$1 = -1

Even add this function calling still not work.
I modified from iscsi_dd.c and removed another link, only keep one iscsi connection. And added timeout in polling.
wk = poll(&fd, 1, 1000);

after iscsi server stopped, in the polling, we get revent=4, and wk always =1, even start iscsi server again, the same things returned.
So I looked into the calling stack, fd=-1, therefor it won't retry.

[sitsofe]

@clicx can you tell us which platform you're using libiscsi on?

[sitsofe]

@clicx Also does https://github.com/sahlberg/libiscsi/blob/e0fd6dc05156f467bffc035be91ea56100b268fa/tests/prog_reconnect_timeout.c work for you?

[clicx]

@sitsofe Linux, I think in qemu built-in libiscsi has same behaviour, while I using an iSCSI target to start a vm, after connection lost for a period of time, the vm will lost disk permanently. But kernel mode iscsi initiator in linux does not, it always keep connection session. So I think if it need any extra configuration on it.

[clicx]

I tested, it seems working in most time, but some time, it happened something:
the event_loop running very fast (means do nothing, just loop), always got: fd=13, wk=1, revent=5
From traffic, it seems continuing logon(successfully) ->logout.