Community
Participate
Working Groups
It's a struggle to fund the infrastructure required to run builds at eclipse. For example see bug 329830, bug 329830 and bug 333594. To raise money for build infrastructure, I propose having sponsored builds. For instance 1) Build message could have an ad included 2) Sponsored build machines 3) Build download pages could have ads etc. Perhaps similar to the ads on the main eclipse.org page. Unconventional? Perhaps. But we need hardware and people to manage it. From my perspective, the eclipse.org infrastructure is not funded appropriately by the board to accommodate the growth of new or existing projects. Everyone is encouraged to run their builds at eclipse.org. One of my plan items for 3.7 is to transition the build to eclipse.org on Hudson. However, to my knowledge, there is absolutely no funding for accommodate this change. (Correct me if wrong :-) The mac Hudson instance that I'm trying to run our tests on is donated from Wayne's basement. It crashes frequently. Last week, I couldn't run any builds on hudson because each time I tried, they failed because the slaves ran out of disk space. New hardware costs money and I understand that budgets are tight. I also understand that the webmasters have a tremendous workload and managing additional hardware adds to their responsibilities, thus the focus on reusing and optimizing existing hardware. However, running builds in a timely manner is also critical to the success of eclipse projects. We are professional open source developers and as such require the appropriate build infrastructure to enable our teams to deliver.
We do have quite a bit of money via Friends of Eclipse. http://wiki.eclipse.org/Friends_of_Eclipse/Funds_Allocation I think the problem would be people to manage the machines?
Chris, I opened a bug for new machines from FoE disbursements and was told no. see https://bugs.eclipse.org/bugs/show_bug.cgi?id=333594#c1
(In reply to comment #2) > Chris, I opened a bug for new machines from FoE disbursements and was told no. > see https://bugs.eclipse.org/bugs/show_bug.cgi?id=333594#c1 Right. Because - as Chris pointed out - the problem is the people costs, not the servers. The operative phrase in that comment is "...to be maintained by the Eclipse Webmasters".
Not to put too fine a point on this, but I don't believe Kim is asking for hardware that would incur additional "people costs". She's just asking for the *existing* infrastructure (e.g. Mac, Hudson disk space,...) to be upgraded to the point that the foundation webmasters can spend *less* time keeping it limping along.
(In reply to comment #4) > Not to put too fine a point on this, but I don't believe Kim is asking for > hardware that would incur additional "people costs". She's just asking for the > *existing* infrastructure (e.g. Mac, Hudson disk space,...) to be upgraded to > the point that the foundation webmasters can spend *less* time keeping it > limping along. If webmaster agrees, then I'm all for use FoE funds for this.
> If webmaster agrees, then I'm all for use FoE funds for this. We agree :) We need disk space for builds.
(In reply to comment #6) > > If webmaster agrees, then I'm all for use FoE funds for this. > > We agree :) We need disk space for builds. Let's go shopping! Should I file a new FoE proposal for this? Or should the webmaster with a more adequate description of funds.
(In reply to comment #7) > Let's go shopping! > > Should I file a new FoE proposal for this? > > Or should the webmaster with a more adequate description of funds. Before we can approve anything, we need a description of the funds required. I don't care where that information comes from, but Webmaster seems like the obvious choice.
> Before we can approve anything, we need a description of the funds required. A disk array, 10 Terabytes or larger in size, to store Hudson Workspace data and other temporary build-related artifacts. How is that? I can spec out the actual hardware.
(In reply to comment #9) > > Before we can approve anything, we need a description of the funds required. > > A disk array, 10 Terabytes or larger in size, to store Hudson Workspace data > and other temporary build-related artifacts. > > How is that? I can spec out the actual hardware. We'll need to get some sense of cost... Do you have as specific unit in mind?
> We'll need to get some sense of cost... > > Do you have as specific unit in mind? I have a specific unit in mind ... $5654 USD + shipping & duty
1. 1 Storform nServ A513 $5654.00 $5654.00 Details: CPU: 1 x Opteron 6128 (2.0GHz, 8-Core, Skt G34, 512KB/Core L2 Cache, 12MB L3 Cache) 80W 45nm RAM: 8GB (8 x 1GB) Operating at 1333MHz Max (DDR3-1333 ECC Unbuffered DIMMs) NIC: Intel 82576 Dual-Port Gigabit Ethernet Controller - Integrated Management: Integrated IPMI 2.0 with KVM and Dedicated LAN Integrated Controller: 6-Port SATA Controller (AMD SP5100) - SAS Controller Required; See PCI Slots SAS 2.0 Expander: Expander provides connectivity to all drives and expansion port (SAS Controller Required) Expansion Port: External SAS 2.0 Connector (24Gb/s, SFF-8088) for JBOD Expansion LP PCIe 2.0 x16 - 1: LSI 9260-4i 6Gb/s SAS/SATA RAID (4-Port Int) with 512MB Cache & BBU LP PCIe 2.0 x8: No Item Selected LP PCIe 2.0 x4 (x8 Slot) - 1: No Item Selected Drive Set: 12 x 1TB Seagate Constellation ES (6Gb/s, 7.2K RPM, 16MB Cache) 3.5" SAS RAID Configuration: RAID 6 with Hot Spare System Volume: 60GB Boot Volume (Carved from RAID Array) Power Supply: Redundant 1200W Power Supply with PMBus - 80 PLUS Gold Certified Rail Kit: Quick-Release Rail Kit for Square Holes, 26.5 - 36.4 inches OS: No Item Selected Warranty: Standard 3-Year Warranty Configured Power: 432 W, 443 VA, 1473 BTU/h, 4.0 Amps (110V), 2.1 Amps (208V) ========================================================================== Total: $5654.00
(In reply to comment #12) > ========================================================================== > Total: $5654.00 Not cheap, but we should see what we can do. Can you open a request against Community->FoE Disbursements and we'll go from there? Also including what the new hardware would be used for.
Bug 336864
(In reply to comment #12) > 1. 1 Storform nServ A513 $5654.00 $5654.00 > > Details: > ... Just thought I'd mention ... I'm sure you already know this ... that what ever system you get or end up with should be able to "delete" files/directories very quickly too ... not just read them. For some reason, on current setup, I've noticed a few times where it takes a very long time to remove a large directory (and, yes, that's using a OS 'rm -fr <directory>') ... seems to take longer to delete it sometimes (like 20 minutes) than it took to write it there (though, I admit, I don't normally watch or measure the time). So ... maybe common knowledge ... but I'm just learning it ... "delete time" is very important for build machines.
> It's a struggle to fund the infrastructure required to run builds at eclipse. It is and always will be. Fortunately, companies like IBM, Google and Oracle are always stepping up to help. More should do so, but that's another thing.
