First off let me say that I am very pleased with ruote. I come from a
BPM background, it was nice to be able to create high level
definitions and not cobble together state machines for my project.
However, I am having some problems with the storage, it appears not
always to update properly. Likely this is a configuration issue on my
end but I'm quickly running out of ideas as to what it could be. I
notice the following symptoms:
1. A newly created process never gets expanded/processed. It is
stuck in the "define" stage, with only 1 expression active. ie. the
process creates but never goes anywhere.
2. Steps (expressions) that are completed, do not get removed from
the storage. For example if I have a sequence with step A and step B,
when step A completes my process will then show *both* A and B as
active.
I have tried using both file system and sequel storage drivers.
Currently the project is in development and we are seeing these issues
with small numbers of users (2-3). We would like to go to production
by end of summer and will be going into beta in June. So I need to
get this problem under control. I don't want to have to babysit ruote
and I would hate to have to rip it out and replace it at this stage.
On Mon, May 21, 2012 at 10:24:26AM -0700, Steve Sexton wrote:
> First off let me say that I am very pleased with ruote. I come from a
> BPM background, it was nice to be able to create high level
> definitions and not cobble together state machines for my project.
Thanks a lot!
> However, I am having some problems with the storage, it appears not
> always to update properly. Likely this is a configuration issue on my
> end but I'm quickly running out of ideas as to what it could be. I
> notice the following symptoms:
> 1. A newly created process never gets expanded/processed. It is
> stuck in the "define" stage, with only 1 expression active. ie. the
> process creates but never goes anywhere.
The way you describe it, it seems that all the processes never go further the
first step. But you mention previously "not always". That's weird.
> 2. Steps (expressions) that are completed, do not get removed from
> the storage. For example if I have a sequence with step A and step B,
> when step A completes my process will then show *both* A and B as
> active.
Is this consistent with the two storage you've been trying?
> I have tried using both file system and sequel storage drivers.
Now time for the questions:
- which platform, OS?
- which Ruby version?
- which version of ruote? (master is recommended)
- single worker architecture?
- sequel against which database? Is the database on the same host as the
workers?
- ...
> Currently the project is in development and we are seeing these issues
> with small numbers of users (2-3). We would like to go to production
> by end of summer and will be going into beta in June. So I need to
> get this problem under control. I don't want to have to babysit ruote
> and I would hate to have to rip it out and replace it at this stage.
The issue you describe tell me "your ruote install is utterly crippled", it
wouldn't be babysitting but life support.
On May 21, 9:09 pm, John Mettraux <jmettr...@gmail.com> wrote:
...*snip*...
> > 1. A newly created process never gets expanded/processed. It is
> > stuck in the "define" stage, with only 1 expression active. ie. the
> > process creates but never goes anywhere.
> The way you describe it, it seems that all the processes never go further the
> first step. But you mention previously "not always". That's weird.
It is a random occurrence. I'd say maybe 90% of the time the
processes behave. When they don't, this is one of the two symptoms.
In this case it acts as if no worker is running (but there is one,
afaik), and no process gets instantiated.
> > 2. Steps (expressions) that are completed, do not get removed from
> > the storage. For example if I have a sequence with step A and step B,
> > when step A completes my process will then show *both* A and B as
> > active.
> Is this consistent with the two storage you've been trying?
Yes. What seems to be happening here is that the step completes but
the record in the documents table hangs around. This is a problem for
me because I have a UI that queries the active steps of a process and
displays them to the user. If I manually delete the record from the
database, all seems to be well.
> > I have tried using both file system and sequel storage drivers.
> Now time for the questions:
> - which platform, OS?
CentOS 5.5 on a x86 VPS
> - which Ruby version?
1.9.2-p290
> - which version of ruote? (master is recommended)
2.2.0, the last official release : )
I suppose I can give the latest stable git version a try, if you think
it will help.
I'm one of those guys that has a preference for only using "official"
releases in production. I'm not religious about it though. As long
as I can configure my app to pull a specific git snapshot, and it
works, that is good enough for me. I just don't like being surprised
by upstream updates when I'm not expecting/ready for them.
> - single worker architecture?
I think this could well be the problem, but I'm not sure which is why
I asked for help. To me, the problem smells like it might be locking,
or a race condition. I have a single worker in each instance of the
rails app. However, I'm running under Apache+Passenger and I have
multiple (currently 6) passenger instances => multiple instances of
rails app => multiple workers, one worker per instance of my rails
app. If that makes sense.
Short-term, scaling back to one worker may be an option. But longer
term I think I will need multiple workers to get the throughput I
need. Also (see below) I see this problem in dev sometimes with
single worker, as well.
> - sequel against which database? Is the database on the same host as the
> workers?
MySQL 5.1. Same host as everything else; workers, apache, passenger,
rails app. It did occur to me this could be a sizing problem. I'm
starting to run out of things to try / places to look, so rather than
guess I thought I'd ask the experts.
I suppose I should add, these problems occur far less often in
development, running under "rails server" (WEBrick) where I *know*
there is only one worker. But they do occur. I get the problems if I
start a lot of processes quickly, close together in time. Our process
definition is still changing, and I have a script I run to re-launch
all the processes with the new definition. If I have, say, 10-15+
definitions getting launched one after the other, I will sometimes see
the above symptoms in dev as well.
I'd say you shouldn't be afraid to use the master, especially in the Bundler
world we now live in. You can peg at the version that works for you.
I'm supposed to release 2.3.0 in June (I'm hoping people on the list we'll
get vocal about it so I get some pressure to release).
If you still encounter the issue, it'd be great if you could "package" it:
"hey John, I'm launching this flow 20 times in a row, and it stalls, we're
using this storage with those settigs, etc". I've seen such issues in the
past and I've done my best to fix them, but sometimes they can re-appear.
Those issues usually don't require much work to reproduce since they're just
jamming the system, a loop is sufficient.
I've quite a number of people using the file system storage and I can think
of an organization in the user list that is using the Sequel storage (they've
been contributing patches (they follow the master version of ruote and
ruote-sequel)).
So I'd say you should try 2.3.0 on master (ruote[ and ruote-sequel]) and
report in detail any weirdness you might find.
