Hacker News Reader

Comment by pointyhat | Hacker News Reader

No they are absolutely abysmal. Stay away for your own sanity.

My FINAL experience with JIRA + Fisheye + Greenhopper + Crucible nearly made me cry. This is on an 8 core Xeon, 12Gb RAM, 15k SAS RAID and Ubuntu 10.04LTS and Sun hotspot VM.

Random failures, null pointer exceptions all over the place that took us out for a couple of days, incredible memory usage, 8xCPUs jammed at 100% due to race conditions, Fisheye only barely just about works in Google Chrome because it's so slow, useless paid support, horridly slow and inconsistent UI, MyISAM tables causing relation failures (!!!!), don't even talk to me about their plugin management - it's a crock of shit.

I couldn't recommend it at all. It's worse than Visual Source Safe.

That's as a JIRA admin. As a JIRA user, I get people saying "can we have Trac back please".

If anyone at Atlassian is listening, please sort your product out. It's not fit for purpose.

View on HN

I've been involved in my company in seven different versions of Jira, and have installed it personally, at other companies, and my own server, and for test purpose no less than 50 times.

The hardware we have installed Jira ranged from a 1 Gigabyte VM with 8 Gigabytes of Hard disk, up to a 2 CPU Dual core server.

90% of the time with mySql as the backend, 10% of the time with HSQL

Operating System has been whatever the LTS variant of Ubuntu, or the more up to date version of RHEL (or CentOS as the case might be)

JRE has always been the most recent version from Sun. We only use fisheye and Crucible internal to our company, but that accounts for 60% of my experience.

In the last four years (and 10 or so versions) - I've never had a random failure or null pointer cause a failure. It's been the most reliable software package I've ever used.

Interesting that we have two completely polar opposite experiences.

My experience is more like yours. I've used Jira a ton of times and any problems we've had using Jira/Greenhopper have been organizational. The only problems I've had with Confluence have been fixed in newer versions.

It is indeed interesting. Thanks for your report. I genuinely wish I'd had your experience.

If you can spare the time, I'd love a more complete report of your issues with our products and support. If you can email me on scott@atlassian.com, I'd love to hear your story and see if there is anything we can do to improve things.

Thanks - I will do that today.

Similar - 1 GB VM, allocated a single core, MySQL backend. Been running for years (although moved from a dedicated machine to a VM a bit over a year ago), solid as a rock. Confluence is running on the same VM.

One of the bits of infrastructure I never have to worry about.

edit: never tried any of their other products however. We use Pulse by Zutubi for continuous builds.

Do you have ~100 developers using it and 1000s of issues? Definitely does not scale.

That's nonsense. I've used it on projects much (MUCH!) bigger than that.

You didn't try and use it with the built in database or something did you?

For a public example, https://issues.apache.org currently has 209560 issues, thousands of users and it works fine.

...until you plug Greenhopper in.

919 Active Developers, 118,460 Issues. MySQL Backend.

#1 rule of Jira is make sure you have enough memory to keep the mySql Database in cache. The attachments don't matter so much.

What I love about the (Lucene?) index is that I can still do a free text search of all 118,000 issues in under 1 second, the same as I was able to when there were only 100 issues. Plus, the 4.x JQL (kind of a simplified Sql Variant) makes it easy to do ludicrously complex searches that still return in basically O(1) time.

DavidU - tell me a bit more about your trouble with Jira Scaling - was it case of the backend system not having enough memory? Jira isn't really disk speed limited, so that's about the only thing that could cause you problems.

Did you get a chance to run the typical performance tools to see if you were CPU, IO, or Memory Bound?

We use Jira in a 1GB VM with Postgres.

No issues.

I get people saying "can we have Trac back please"

People actually like Trac? I thought people used it because it is better than Bugzilla and is free.

Yes - lots of people like it. It's actually an incredible piece of software which allows you to build what you want on it.

Any chance I could chat with you about your experience with our support? If so, drop me an e-mail? donna at atlassian.com

"useless paid support"

Sorry if you feel the support is useless. We try our hardest but that's your opinion and you're entitled to it.

However, saying it's "paid" is flat-out wrong. No customer ever pays for support from Atlassian. It's included in the license fee. Free support to current customers is the only type of support we offer.

I'm talking about the "maintenance agreement".

The funny part is trac isn't that fast either. It's overall a superior tool for the job, and loads simpler in every respect, but I get a healthy amount of delay on each and every action I perform in most trac deployments.

I keep meaning to take time out one of these weekends to profile and determine how to mitigate this, but it's just not a priority.

Switch to the postgres backend. The default sqlite one can get quite slow in high concurrency situations, although I've actually had 50 peple running quite happily off the sqlite backend.

Present concurrency is 0, it's just slow. It's not SQLite unless there is pathological behavior because I've demoed ideas to HN traffic with SQLite before and it performed better than a vanilla MySQL deployment. (Not saying this is advisable, just that SQLite isn't at fault unless Trac is doing something stupid.)

Are you positive switching to the postgres backend will help when there's almost no concurrency and the slowness is persistent and deterministic?

All I'm saying is that it worked for me.

How are you hosting it? I'm using Apache 2.2 with mod_wsgi on Ubuntu 10.04LTS. Don't use mod_python - it's terrible.

Reply on news.ycombinator.com