Internet auction giant eBay said it has restored service after a sequence of
events starting with a hardware failure that took the site off-line for
almost 11 hours yesterday.
The company said that the failure, which lasted until 10 p.m. PST, “was the
result of a series of failures that affected both our primary and back-up
systems. This was compounded by a decision to delay the replacement of
certain hardware components in an effort to avoid disrupting service during
the busy holiday season.”
eBay said the site outage was not expected to have any material financial
impact. In fact, Goldman, Sachs analysts said that they “believe the revenue
impact is less than $500,000 in credits to sellers.”
GS said it continues to advise clients to add to market positions in the low
$30 range. eBay stock actually went up yesterday as the overall market
rocketed higher, closing at $39.35, up more than $9. It was down almost $3 at
mid-morning today.
The site has had several failures over the past few years, including some
caused by denial of service spam attacks. eBay experienced one of its largest
outages on June 10, 1999, which lasted over 22 hours and resulted in the
company paying users approximately $3.9 million in credited fees, according
to Goldman Sachs.
eBay also issued a formal apology, saying: “We apologize to our community for
this lengthy interruption of service.” eBay said it would extend all listings
by 24 hours and credit related fees.
The outage was due to a series of unrelated issues that began with a hardware
failure on eBay’s back-up system. As the company restarted the back-up
system, another problem developed in the storage system shared by the primary
and back-up systems, effectively bringing down the site.
A third system was then brought up, and it functioned effectively for about
40 minutes before a database problem caused this system to also become
unavailable, the company said.
Interestingly, eBay had this to say in a statement on the system failure:
“While we have known for a while that a potential problem existed in our
shared disk hardware, we chose to delay the recommended upgrade because we
had developed a series of work-arounds that had previously proven effective.”
The company said it will upgrade hardware elements in the next few weeks,
which will require scheduled downtime of approximately six hours. “In
addition, we have already embarked on a longer-term program to distribute the
database to many separate servers that will isolate any failure to a limited
part of the site,” the company said.
