Unsolved
This post is more than 5 years old
4 Posts
0
28528
October 21st, 2010 04:00
Dropped packets on 2824 with iSCSI
hi,
my dropped packets counter keeps rising on initiator ports (ESX4i sw initiator).
my targets are intel gbit cards (PE T710 servers)
my initatirs are intel gbit cards (PE R610 servers)
i am running just two targets (mirrored san) and three initiators. not much load.
target ports are clean. just the esx initators are producing dopped packets.
i have set up the esx hosts as described in the iscsi config guide fuer esx4.1
cabling is cat5e SFP 3m just within the rack.
switches and servers got delivered around june this year. so pretty new.
the SAN overall is doing fine except for some storage lag which doesn't influence production.
any clues?
regards
kt
0 events found
No Events found!


bh1633
909 Posts
0
October 21st, 2010 16:00
enable flow control on switch and on all NICs. This will eliminate dropped packets.
KThost
4 Posts
0
October 25th, 2010 06:00
flow control on the switch is enabled, and is enabled by default on ESX hosts.
so it should be on on all related components.
ideas?
bh1633
909 Posts
0
October 25th, 2010 07:00
I would examine the counters on the NICS and switches closely. See if there are any other clues like crc errors.
How are you reading the "dropped packets" counter on your servers?
bh1633
909 Posts
0
October 25th, 2010 08:00
Please post the switch counters for the connected ports.
show interfaces counters
To be honest, the percentage of dropped packets is so low, you are unlikely to notice any performance problems.
Is there any significant non- iscsi traffic on the switch? If so, you should consider putting the iSCSI traffic in its own vlan.
KThost
4 Posts
0
October 25th, 2010 08:00
i am checking the RMON statistics. no CRC or other errors are visible.
g1: (target)
Dropped packets 0
Received Bytes (Octets) 2235772662
Received Packets 1344907550
Broadcast Packets Received 13963
Multicast Packets Received 1134
CRC&Align Errors 0
Undersize Packets 0
Oversize Packets 0
Fragments 0
Jabbers 0
Collisions 0
g2: (target)
Dropped packets 0
Received Bytes (Octets) 277601478
Received Packets 1841326852
Broadcast Packets Received 384
Multicast Packets Received 627
CRC&Align Errors 0
Undersize Packets 0
Oversize Packets 0
Fragments 0
Jabbers 0
Collisions 0
g3: (initator ESX host)
Dropped packets 132
Received Bytes (Octets) 4057227261
Received Packets 23629315
Broadcast Packets Received 166
Multicast Packets Received 6
CRC&Align Errors 0
Undersize Packets 0
Oversize Packets 0
Fragments 0
Jabbers 0
Collisions 0
g5 (initiator ESX host)
Dropped packets 205
Received Bytes (Octets) 2179935489
Received Packets 778684903
Broadcast Packets Received 580
Multicast Packets Received 0
CRC&Align Errors 0
Undersize Packets 0
Oversize Packets 0
Fragments 0
Jabbers 0
Collisions 0
g21: (initiator win2008r2 host)
Dropped packets 0
Received Bytes (Octets) 3853829328
Received Packets 379100714
Broadcast Packets Received 13335
Multicast Packets Received 524
CRC&Align Errors 0
Undersize Packets 0
Oversize Packets 0
Fragments 0
Jabbers 0
Collisions 0
thanx for you time.
kt
KThost
4 Posts
0
October 26th, 2010 09:00
the cli doesn't offer "show interfaces counters"
the counters are from the RMON webinterface status.
there is no other traffic than iscsi traffic on the switch.
bh1633
909 Posts
0
October 26th, 2010 12:00
Do the counters on the servers show anything?
Dropped packets counter indicates that the device dropped some packets because it ran out of resources. Enabling flow control on your network (all switches and NICs) should prevent this.
If you can find the controls in your OSes, you might try reducing the TCP window size until there are no dropped packets.
I am out of ideas other than try another switch and see if you get same results.