Monitoring:
Live data sources:
isi statistics is point in time look
SNMP data - tells whether cluster is out of normal situation, cache, memory
cluster events - like drive failures , boot failures
Exporting Historic data
isi_stats_d
InsightIQ history
CELOG database:
Logs
efs.gmp.group
isi statistics system
pstat
client
protocl
query
isi statistics system --nodes --top teslls most active node
cluster wide and protocol data isi statistics pstat
isi statistics pstat --protocol=smb2 --top
nlm: network lock manager for nfs version 2 and 3
lsass_in and lsass_out authentication
isi statistics client displays most active clients accessing cluster for each protocol types
isi statistics client --orderby=ops --top
isi statistics protocol --totalby=proto,class,op,node --top
BDP: Bandwidth delay product calculates how much bandwidth available on the network. Calculating bandwidth (RFC 1323)
64KB/RTT (round trip time) = speed of NIC HW
ex 64kB/0.5 milliseconds) = 125 MBPS
Network performance measurements tools
iperf: measures TCP and UDP streaming throughput between endpoints
ping - measures RTT, latency on network wire
wireshark/Tshark allows comprehensive examination of numerious protocols, live/offline capture analysis
tcpdump: provides ability to view packets using predefined fileters, or by capturing everything for post capture filtering (wireshark or tshark)
Iperf usage:
Client: iperft -c <iperf server IP> -t 20
Server: iperf -s
workload I/O operations
Name space operations: mkdir,getattr,readdir,lookup
to see stats for reads,writes, namespace reads, and namespace writes:
isi statistics protocol --classes=read,write,namespace_read,nameshpace_write
how many operations are queued up for each disk
isi_for_array -s sysctl hw.iosched |grep total_inqueue
No comments:
Post a Comment