Project

General

Profile

Actions

Support #2548

closed

suricata flow management in 10gbs environment

Added by sandy sun almost 6 years ago. Updated almost 6 years ago.

Status:
Closed
Priority:
Normal
Assignee:
-
Affected Versions:
Label:

Description

I have a problem in using suricata as ids mode in idc 10gbps environment.
I found finallly flow recycler cleanup flows slower than flows created. this lead to flow spare queue decrease to zero and recycle queue very large。especially when enter emergency mode,flow manager module schedule more frequent than flow recycler module。

I decreased flow-timeout conf,configed more than one flow recycler modules, but not effictive as assumed。
flow:
memcap: 2gb
hash-size: 1048576
prealloc: 1048576
emergency-recovery: 30
managers: 1 # default to one flow manager
recyclers: 6 # default to one flow recycler thread

flow-timeouts:

default:
new: 30
established: 200
closed: 5
bypassed: 100
emergency-new: 10
emergency-established: 100
emergency-closed: 0
emergency-bypassed: 50
tcp:
new: 20
established: 200
closed: 5
bypassed: 60
emergency-new: 5
emergency-established: 100
emergency-closed: 0
emergency-bypassed: 50
stream:
memcap: 2gb
checksum-validation: yes # reject wrong csums
prealloc-sessions: 65536 # sessions prealloc'd per stream thread
inline: no # auto will use inline mode in IPS mode, yes or no set it statically
bypass: yes
reassembly:
memcap: 10gb
depth: 512kb # reassemble 1mb into a stream
toserver-chunk-size: 2560
toclient-chunk-size: 2560
randomize-chunk-size: yes
#randomize-chunk-range: 10
#raw: yes
segment-prealloc: 65536
#check-overlap-different-data: true

The stats as below:
------------------------------------------------------------------------------------
Date: 7/23/2018 -- 16:05:28 (uptime: 0d, 00h 59m 29s)
------------------------------------------------------------------------------------
Counter | TM Name | Value
------------------------------------------------------------------------------------
flow_mgr.closed_pruned | FM#01 | 77801406
flow_mgr.new_pruned | FM#01 | 14559621
flow_mgr.est_pruned | FM#01 | 17102653
flow_mgr.bypassed_pruned | FM#01 | 910
flow.spare | FM#01 | 142790
flow.emerg_mode_entered | FM#01 | 1
flow.tcp_reuse | FM#01 | 2266271
flow_mgr.flows_checked | FM#01 | 9386320
flow_mgr.flows_notimeout | FM#01 | 3213988
flow_mgr.flows_timeout | FM#01 | 6172332
flow_mgr.flows_timeout_inuse | FM#01 | 1446235
flow_mgr.flows_removed | FM#01 | 4726097
flow_mgr.rows_checked | FM#01 | 1048576
flow_mgr.rows_skipped | FM#01 | 4
flow_mgr.rows_empty | FM#01 | 341
flow_mgr.rows_maxlen | FM#01 | 67
tcp.memuse | Global | 1804674600
tcp.reassembly_memuse | Global | 10737418208
dns.memuse | Global | 1398589
http.memuse | Global | 2783051133
flow.memuse | Global | 2147483488

Is there any method or slolution for this problem?


Files

stats.log (94.7 KB) stats.log sandy sun, 07/23/2018 09:48 AM
stats.log (3.47 MB) stats.log sandy sun, 07/24/2018 12:37 PM
Actions #1

Updated by Peter Manev almost 6 years ago

Which Suri version are you using?
Can you please post a full record of the last update in stats.log?

Actions #2

Updated by sandy sun almost 6 years ago

Peter Manev wrote:

Which Suri version are you using?
Can you please post a full record of the last update in stats.log?

I use 4.0.0 version
23/7/2018 -- 15:05:51 - <Notice> - This is Suricata version 4.0.0 RELEASE

Can you please post a full record of the last update in stats.log ====
I restart again, the full stats record file in stats.log attachments.thanks.
do you need more verbose stats items?

Actions #3

Updated by Peter Manev almost 6 years ago

Can you please try with a supported version of Suri - 4.0.5 or it would be even better with 4.1RC1.
In the attached stats.log - it seems it has run for just 6 minutes and there is no emergency mode entered as you describe previously. Also no decoder/alert records in stats.log.

Actions #4

Updated by sandy sun almost 6 years ago

Peter Manev wrote:

Can you please try with a supported version of Suri - 4.0.5 or it would be even better with 4.1RC1.
In the attached stats.log - it seems it has run for just 6 minutes and there is no emergency mode entered as you describe previously. Also no decoder/alert records in stats.log.

I may use the 4.0.5 or 4.1RC1 version.
I upload a recent stats.log with decoder & app counters from start to enter emergency mode.
Last time without these counters because i forgot add StatsSyncCountersIfSignalled in my own acq pkt loop.
I want to known whether the stream/reassemble memory configed is not enough or the flow recycler is slower than actual need.
Thks.

Actions #5

Updated by sandy sun almost 6 years ago

hi all,I have solved this problem;
I forgot call RunModeInitialize in register new runmode, lead to fm/fr set affinity failed.
finally, how can i close this problem? thks

Actions #6

Updated by Peter Manev almost 6 years ago

  • Status changed from New to Closed

Closing upon request.
From your last comment - I am not really sure how the problem was solved (or if custom code additions were used).

Actions #7

Updated by Peter Manev almost 6 years ago

  • Assignee deleted (Victor Julien)
Actions #8

Updated by Victor Julien almost 6 years ago

  • Tracker changed from Optimization to Support
  • Effort deleted (medium)
Actions

Also available in: Atom PDF