Ragged Flash: A Smooth Solution to a Thorny Problem
June 15, 2022 Alex Woodie
There are many benefits to using IBM’s FlashCopy services, which delivers a nearly instantaneous copy of a customer’s IBM i environment. But there are also challenges, such as when a FlashCopy is paired with an IPL, which had the habit in busy environments of throwing the source and the copy out of synch. IBM has addressed this problem with “ragged flash.”
Ragged flash is one of the new database features delivered in IBM i 7.5 (it’s not available in 7.4). Scott Forstie, the database architect for IBM i, identified the ragged flash as one of his most favorite new database features. During the recent POWERUp 2022 conference in New Orleans, Forstie explained how ragged flash works and why it’s so cool.
“The behavior before 7.5 was, after you flash, what do you want to do? You want to IPL so probably you can do a backup,” Forstie said. “So the act of IPL, you automatically roll back any active database transactions, so your copy now is different than your source.”
It’s great that FlashCopy gives IBM i shops the ability to snapshot their entire system or iASP in an instant to perform necessary maintenance, such as running a backup and then restarting the system to apply updates. But that speed should not come at the cost of the accuracy of their transactions, which is what was happening.
There was one workaround, Forstie said, but it was a hard one to execute in the real world. “How our clients have been dealing with this is they ask their business unit ‘Hey everybody, can you stop for a second so we can do a copy?’ And nobody does!” he said. “How do you quiesce activity? It’s not easy.”
The promise of 24/7 operations is now here, and many IBM i shops simply don’t have the spare bandwidth to log off their systems for even a few minutes, Forstie said.
“There were a couple of commands in the operating system where you can try to enforce [quiet time],” Forstie says. “You can say ‘Stop ASP activity and quiesce! Quiesce! But in practice, everybody I’ve talk to said that’s practically impossible, because the promise of 24/7 is among us. People are active all the time. You can’t find that quiet time.”
Luckily, IBM is now providing them a solution to this dilemma with ragged flash. Forstie explained how ragged flash works:
“The client on the flash can tell us, don’t roll back the pending transitions. So it leaves it in an unusable state,” he said. “But if you bring over the journal receivers, you can apply or remove them. You can move forward or you can move backward, so [this update delivers] operational control of backups using flash.”
There’s a bit more to it, and you can read more of the gory details in its entry on ragged flash in an IBM Support article on the topic.
In a way, ragged flash is a modern twist on an old IBM i standby, Save While Active. That old feature could allow users to continue accessing database objects while it was being backed up, with a short quiesce required on the backend to tidy up any loose ends.
While traditional save restores to tape have their place, newer technologies like FlashCopy work at a different layer of the stack, Forstie said.
“One of the nice thing about FlashCopy is it’s at a higher level,” Forstie said. “It’s sector level. You’re not dealing with objects. You’re dealing with an entire IBM i, and it’s very fast. It has some distinct advantages.”
The vexing problem of choosing between speed or completeness with FlashCopy is no longer the problem it once was, and that makes Forstie a happy camper. “If you were a client who’s utilizing FlashCopy services, you’d say ‘Oh wow, you got that figured out!’” he said.