Thursday, 18 December 2014

System Slowdowns and CPU running 100% - Mitel 5000 / MiVoice Office 250

One of the most frequent complaints we get is that of system slowdown. So the user may report that when the phone rings and they lift the handset, that it takes a couple of seconds to answer, or that the display on the phone changes slowly, or when they dial digits they take a few seconds to register.

This is usually caused by the CPU being run at or close to 100%. The avaerage system usually ticks over averaging 20 - 40%, so seeing the processor flat out for a prolonged period rings alarm bells.

Most Common Causes We Have Found


  1. Sip Hackers - The number one cause. 90% of issues turn out to be this. All versions of the 5000 have a SIP port 5060 open by default, even the versions that did not support SIP trunks or extensions had a SIP port. If this is open to the internet without a firewall filtering by source IP then your system will get bombarded by call requests, sometimes thousands a second. Even though the calls don't get placed, the effort of responding to them causes the system to slow down.  You can see the messages if you turn on SIP logging. Either lock down SIP using the firewall or turn of the SIP port in IP Settings / advanced
  2. SSH Hackers - The same thing will happen if you leave port 22 open to the internet. Some Bot will come along and bombard the system with login messaages trying to find the password. You can see the fail messages in the syslog.  Either lock down or turn off SSH.
  3. Voice-mail to email - Whilst it usually does not cause too much of an issue, sometimes the site are heavy voice-mail users and have extended the recording time available per message, or worse yet are using the embedded voice-mail for record-a-call.  The turning of the voice-mail into an mp3 then emailing it can cause the processor to show heavy loading. I tested this once with a 20 minute call recording and it took the processor an hour of running flat out befoire it calmed down.
  4. SSL Problems - Getting a bit rarer now, but we have found a site where an SSL error was being shown in the exception log. Using the web interface of the system the certificate was regenerated and the web server restarted. This must have unlocked an internal process as immediately the processor use dropped.
  5. ISDN Errors - Whilst Open reach are usually insistent there is no error, we have seen circumstances where ISDN can flood the system with error messages causing the system to slow down as it attempts to handle them all and record them in the logs.
  6. Over use of G729 - One of the most common mistakes engineers make which causes load on the CPU and can affect voice quality is to use G729 where it is not necessary. G729 should only be used where there is a real issue with having enough bandwidth for G711. On the LAN where there is bags of bandwidth it is never needed. Even on broadband you may not need it. G729 compresses the speech, which while saving bandwidth, requires more resources on the CPU to do the compression/decompression, so whenever I see a high load on a CPU I look to change to G711 and this usually has a visible effect on CPU use.
I hope this may help someone in the future solve a problem!

John Rogers

No comments:

Post a Comment

Please keep your comments polite and free from bad language. Thanks

LetsEncrypt failure on Draytek Routers

 We like to use Draytek Routers on our installs as they are easy to configure and tend to be reliable. As with all things these days https s...