Open NickeZ opened 5 years ago
Solved it with
systemctl stop archiver
find /home/archiver/ -size 0 -delete
systemctl start archiver
And now it looks better in the table:
PV_NAME | Being archived | appliance0 | true | true | 1.0 | Dez/05/2018 12:59:32 +01:00
But why is AA creating empty zipfiles?
Are you by any chance using compress=true? http://slacmshankar.github.io/epicsarchiver_docs/api/edu/stanford/slac/archiverappliance/PlainPB/PlainPBStoragePlugin.html
The support for compression is not great; it's been experimental for quite some time and probably ought to be removed. The .zip file format does not lend itself to this kind of storage easily. I would not use compression; the possibility of data loss is quite real in this case.
We should redo the support for compression at some point in time.
Yes we use compression. Without it we need several orders of magnitude more disk space which isn't feasible..
You'd be the first team to try it out in production. Good luck. Did you check out ZFS compression or equivalent? This may have better results and will probably get you significantly better performance.
When you put it that way I might reconsider.. we have a very small system with only a single ioc and a few thousand pvs. What kind of failures should I look out for?
What kind of failures should I look out for?
Not sure; we made the mistake of using the .zip format for compression with the idea that this would give us the ability to compress and store multiple files in the same zip file. However, the .zip file format is not suited for append workloads; which is the main kind of workload we have. When we add content to the .zip file, the existing file is copied and a new file with the newly added content is generated. There are multiple potential points of failure in this process; not to mention the terrible performance.
And since compression is quite readily available at the file system level; this feature has not be pursued as much. I have not had the time to refactor the code to pull this feature out; I'll probably do this soon.
Hey,
I get the following exception in
retrieval_catalina.err
when the PV does not yet contain any data.It looks something like this in the table: