Cisco has launched a replacement program for one of its B-series Unified Computing System (UCS) blade servers after revealing a faulty transistor could cause the unit to overheat and damage nearby blades.
In a field notice issued last week, Cisco said it was contacting customers to replace the transistor in its UCS B440 Blade Servers with modified hardware.
The notice warned that failure of the server's metal-oxide-semiconductor field-effect transistor (MOSFET) can cause it to overheat and emit a flash that could lead to "complete board failure".
"In extreme circumstances it could affect the other blades in the chassis by disrupting power flow," it reported.
Efforts to contact customers began after a thermal incident occurred at a customer site. It did not say when that incident occurred. However a firmware update was released in July last year in response to an incident that occurred at one customer site.
A second incident triggered modification of the hardware and the replacement program.
Cisco told CRN.com the thermal event had occurred at two customer sites, adding that it was "proactively managing this situation" by keeping customers informed of when the modified hardware became available.
"A hardware replacement program has been launched and a modified UCS B440 Blade Server will be announced when it becomes available," Cisco said in the notice.
It was also discussing the option of alternative UCS solutions for customers.
No other UCS hardware was affected, Cisco reported.
Cisco has provided workaround details that will help minimise the risk of damage if a transistor fails, including upgrading to the latest UCS Blade Management Controller Software.