Outages plague eBay again
Following a five-hour outage this morning, technically troubled eBay's bidding system is running again, but not without glitches.
Today's problems, plus sluggish performance yesterday, are not related to the 22-hour outage on June 10-11, according to eBay spokesman Kevin Pursglove. Postings on eBay's site today and yesterday point to problems with eBay's CGI server pool.
While the site is back up, not all features are working. At 11:15 a.m. PT today, eBay said it had disabled its email servers that send status reports to users when an auction ends and when bids are received.
In addition, updates to listings and searches have been delayed.
Pursglove blamed yesterday's problems on a data transfer that began freezing the auction listings area, making the system run at a crawl. Pursglove said the site was accessible yesterday for all but half an hour.
The recurrence of problems at eBay, one of the Internet's busiest sites, points to the fragility of the hardware and software that run sensitive e-commerce and infrastructure systems that millions of Internet users rely on. Just yesterday, free email service Hotmail experienced problems, and users of Excite@Home have complained they're not getting the speed they pay for.
Experts say such problems cannot be fixed easily.
"Largely, not anyone is ready for the demand," said Cormic Foster, analyst with Jupiter Communications. "[These companies] are cursed by their own success. I don't think you'll ever stop seeing these things happening, at least while the Internet is still in growth mode."
With "get big fast" as the main goal for many Web companies, scalability problems are likely to persist. However, not all problems are related to heavy usage.
eBay, for example, has insisted through its numerous outages in the last six months that rapidly growing traffic is not to blame. Now it's tempering that statement, saying it really doesn't know what started the problems this month.
But in the big picture, more traffic means more servers, more routers, and more bandwidth just to keep up. But Jupiter's Foster says adding hardware often isn't enough.
"Software often is the thing that fails, so you really need to test your software applications," he said. That becomes a significant challenge with the frequent updates of site design and features that are common with the largest Web sites.
One aspect of eBay's continuing problem is that it has not had a backup system that can be used when the main system fails, a flaw eBay's Pursglove said the company was working to repair. Had the June 10-11 problems come a few weeks later, the backup system might have been ready, he said. But eBay's backup system is under construction.
In the major eBay outages of earlier this month, the first cause identified publicly was "Sun software," in a press release from both Sun Microsystems and eBay.
Subsequently, however, Sun chief executive Scott McNealy said eBay failed to install a software patch for the Sun Solaris operating system software, a remedy that others at Sun claimed could have prevented the outage.
eBay and its vendors haven't pinpointed the precise cause of the June 10-11 problems, Pursglove said, but sources close to the situation suggest something triggered a chain of events that accelerated and eventually put the whole system out of service. Ebay's infrastructure includes Oracle database software as well as software from Veritas for managing applications to continuously run around the clock.